Our latest AI Pulse survey, taken by listeners of The Artificial Intelligence Show, shows a significant majority of professionals (two-thirds) view AI as an immediate or near-term career threat, even as they rapidly integrate it into their daily workflows. Continue reading “67% of Professionals See AI as a Near-Term or Immediate Job Threat”
Generative AI refers to algorithms and models that can create new content, designs, or predictions by recognising patterns from large amounts of data. At their core, these models are trained by exposing them to vast datasets, allowing them to pick up statistical patterns and relationships. Once trained, they can be prompted with a seed input and generate contextually relevant output, often in a way that feels creative or human-like. Leading models today include OpenAI’s GPT‑4o, Anthropic’s Claude‑3, Google’s Gemini, Meta’s Llama 3 and Mistral’s Mixtral.
From a business perspective, generative AI represents not just a technological novelty but a transformative force: it can automate tasks, augment human creativity and unlock new revenue streams. Adoption doubled to 65 % of companies by early 2024, and 92 % of Fortune 500 firms had begun using it. Investments deliver outsized gains—every dollar spent on generative AI yields about $3.7 in value, with financial services seeing ROI as high as 4.2×. Analysts project the generative AI market to reach $644 billion by 2025.
Clarifai integrates both proprietary and open‑source foundation models (from OpenAI, Cohere, Anthropic, GPT‑Neo, BERT, Stable Diffusion and others) into a single platform. Beyond model access, Clarifai provides data augmentation, content generation, vector store and prompt library modules, enabling enterprises to tailor generative solutions while maintaining privacy and performance through features like local runners.
Q: What does generative AI do that traditional AI cannot?
A: Generative AI creates new data—synthetic images, text, code or audio—by learning patterns from training data, whereas traditional AI typically classifies or predicts based on known patterns.
Q: Why is adoption accelerating?
A: Wide availability of foundation models, lower compute costs and platforms like Clarifai have made deployment easier. Adoption doubled to 65 % of firms by early 2024, and ROI per dollar invested averages $3.7.
Q: How does Clarifai fit in?
A: Clarifai integrates multiple foundation models, data labeling, model training, AI workflows and vector databases into one ecosystem, letting organisations deploy generative AI securely and at scale.
Generative AI’s adoption curve is remarkable. In just a year, the share of enterprises experimenting with generative AI jumped to 65 %, and 71 % now use it in at least one business function. Sector‑specific adoption rates show where the technology has immediate traction: healthcare (47 %), financial services (63 %), media/entertainment (69 %) and education (55 %).
This momentum translates to significant value. McKinsey projects generative AI could contribute $2.6–$4.4 trillion annually, with 75 % of the benefit concentrated in customer operations, marketing & sales, software engineering and R&D. Process automation is a major driver: tasks like drafting emails and writing code can be largely handled by AI, freeing people for higher‑value work. Some businesses already report savings of 4–9 hours per employee per week.
From a customer‑experience standpoint, generative AI is revolutionising service delivery. Surveys show 70 % of customer‑experience leaders plan full integration by 2026, and generative chat reduces service costs while improving customer effort scores by 57 %. Marketing departments are ahead of the curve: 92 % plan to invest in generative AI, with 78 % already using AI for content creation/SEO.
The next frontier is multimodality—models that process text, image, audio and video simultaneously. The multimodal AI market, valued at around $1.2 billion in 2023, is projected to grow at more than 30 % annually. Alongside closed models, open‑source models like Meta’s Llama‑3.1 and Mistral’s Mixtral deliver performance nearing proprietary models. Clarifai supports these open models, letting enterprises fine‑tune them with proprietary data while retaining control.
Generative AI is moving beyond passive chatbots to agentic systems that can orchestrate tasks across multiple tools without constant user input. They can triage support tickets, draft reports, recommend actions and even execute workflows within Clarifai’s Mesh AI orchestration engine.
Large language models sometimes hallucinate; to mitigate this, more applications combine LLMs with retrieval‑augmented generation (RAG). RAG uses vector databases to index documents, enabling the model to fetch factual context before generating an answer. About 28 % of organisations already use vector databases and another 32 % plan to adopt them. Clarifai’s Vector Store provides ready‑to‑use vector search across unstructured text, images and video.
As privacy regulations tighten and data scarcity persists, generative AI becomes a tool for synthetic data generation. Clarifai’s synthetic data module splits large documents into sections, stores them in a vector database and allows users to query them securely . This approach overcomes input length limitations and reduces the risk of exposing sensitive information.
Governments are drafting AI‑specific regulations focusing on privacy, fairness and transparency. Meanwhile, nearly 45 % of businesses report a shortage of AI skills. Companies must invest in upskilling and adopt human-in-the-loop workflows to ensure safe deployment.
Developers use AI to handle boilerplate tasks such as writing APIs, generating tests and translating languages. Tools like GitHub Copilot suggest blocks of code based on comments. Productivity increases significantly, but human review remains essential to avoid security vulnerabilities.
Generative models also assist with devops and MLOps. Clarifai’s Mesh engine orchestrates complex pipelines: a natural-language specification triggers an LLM to generate skeleton code, uses the Vector Store to retrieve relevant API documentation and connects to testing tools. By automating code reviews and test generation, generative AI reduces human error but still requires engineers to validate outputs.
Clarifai’s vector store indexes entire codebases, enabling retrieval-augmented generation. Developers build chat-based assistants that answer questions like “How do I connect to our payment API?” by retrieving relevant code samples. Clarifai’s prompt library provides templates for code tasks, ensuring consistent style. Local runners allow models to run within private environments, protecting proprietary code.
Generative AI upgrades chatbots into empathetic agents that understand context, detect sentiment and provide human-like responses. They handle multi-turn conversations and even recognise images or video for diagnostics. Contact centres benefit through automation: AI triages tickets, drafts responses and summarises calls. Marketing departments lead adoption, with 78 % using AI for content creation/SEO.
Clarifai’s Language Chains structure workflows that combine retrieval from knowledge bases with generative reasoning. For example, a Clarifai-powered support agent can identify intent, query the Vector Store for relevant documentation, draft a personalised response and summarise for human review. The stuffing chain pattern splits large knowledge bases into segments, ensuring the model considers all relevant documents. Local runners ensure sensitive data stays on-premises.
Generative AI tailors education to each learner by analysing performance data and creating customised content. Tutors answer questions in natural language, generate examples and adjust difficulty levels. Beyond text, multimodal AI creates interactive simulations, enabling students to experiment safely.
Clarifai accelerates educational innovation through:
Generative AI simulates market scenarios, generates synthetic data for stress tests and drafts regulatory documents. Chatbots answer investment queries, while analysts use AI to summarise earnings calls. Fraud detection benefits from models that identify anomalies in transaction data, flagging potential fraud. AI also creates personalised investment recommendations based on client profiles.
Clarifai enables finance teams to:
Generative AI models normal and abnormal behaviours from historical data and simulates new fraud patterns, enabling proactive detection. Synthetic fraud scenarios help models learn to recognise novel attacks. In regulatory compliance, AI automates document analysis and identity verification.
Clarifai equips risk teams with:
Generative AI tools produce high‑quality images from text prompts, transform scripts into animated sequences and compose music. It enables rapid ideation, helping non-specialists create professional content. Generative AI also scales marketing assets and summarises video libraries for easy discovery.
Clarifai bridges content creation and management by automatically labelling objects and scenes, feeding them into the Vector Store for retrieval. It supports fine‑tuning of open-source models with proprietary data and provides video intelligence for tagging and summarising scenes.
Generative AI enhances images for early detection, summarises patient notes and designs novel drug molecules. Multi‑modal models integrate EHRs, imaging and genetics to propose personalised treatments. However, strict privacy and ethics requirements govern deployment.
Clarifai offers local runners to ensure PHI remains secure, synthetic data modules to generate de‑identified datasets, vector search for EHR and literature retrieval and support for fine‑tuning open-source models.
Generative AI extracts skills from résumés, matches candidates to job requirements, schedules interviews and drafts offer letters. For development, AI creates tailored learning pathways and summarises reviews.
Clarifai provides model training and evaluation, vector store for candidate and job data, synthetic data for training diversity simulations and local runners for secure processing.
Generative AI drafts policy documents, calculates premiums, summarises claims and flags anomalies. Chatbots handle customer inquiries. For underwriting, AI synthesises data to estimate risk; synthetic data trains models to detect fraud.
Clarifai uses local runners to protect sensitive data, vector store to organise policy documents, synthetic data module for training fraud models and Mesh for workflow orchestration.
Generative AI speeds contract creation and review, summarises cases and performs e‑discovery. For compliance, AI monitors regulatory changes and generates reports.
Clarifai offers vector store for statutes and cases, local runners for secure processing, prompt templates for drafting and a PDF import module for long documents.
Generative design tools propose numerous novel designs meeting specified objectives. AI accelerates material discovery and optimises 3D printing. Digital twins integrate sensors and AI to simulate product behaviour.
Clarifai offers Scribe for data labeling, Enlight for training, Mesh for orchestration and Spacetime for managing design data.
Generative AI analyses project data to predict timelines and optimise resources. It auto-generates status reports and summarises meetings. Agentic systems can adjust schedules autonomously, subject to human approval.
Clarifai provides:
Copywriting assistants generate tailored emails, posts and landing pages. AI analyses user data to recommend resonant content. For SEO, generative models suggest keywords and write optimised meta descriptions. Generative AI also creates video ads, interactive demos and 3D product renders, enabling smaller brands to compete.
Clarifai amplifies marketing through:
Generative AI simulates demand fluctuations and supply disruptions, creating contingency plans. It recommends optimal reorder points and adjusts routes in real-time based on traffic and weather. AI drafts procurement contracts and negotiates terms via chatbots.
Clarifai provides vector store for unstructured supply data, synthetic data modules for simulations, local runners for secure data and Mesh for unified decision engines.
Synthetic data provides privacy-friendly datasets when real data is scarce or sensitive. It trains models, stress-tests systems and aids fairness auditing. In computer vision, synthetic images improve robustness; in language tasks, synthetic documents cover edge cases.
Clarifai’s module splits documents into chunks, embeds them and stores them in a vector database. This enables targeted retrieval and privacy. It also supports fine‑tuning generative models with synthetic data and allows secure local deployment.
Generative AI has transitioned from novelty to necessity. Adoption stats show a majority of enterprises experimenting with generative AI, and investment is skyrocketing. Yet the full potential remains untapped; many projects stall due to data challenges, and ethical concerns must be addressed.
As we look to 2026, multimodal models will merge text, vision and audio, enabling richer interactions. Open-source models will democratise access. Agentic AI will take on complex tasks autonomously, while RAG and vector databases will ground models in factual context. Synthetic data will alleviate privacy concerns. Regulation will ensure responsible deployment.
Success hinges on data quality and the ability to fine-tune general models with domain-specific knowledge. Platforms like Clarifai—integrating foundation models, vector stores, prompt libraries and orchestration tools—offer a comprehensive solution. The future of generative AI lies not just in technology but in responsible, creative and collaborative implementation.
Traditional AI systems typically classify or predict based on existing patterns. Generative AI, by contrast, creates new content—texts, images, music or code—by learning underlying patterns from training data.
Begin with clear use cases—such as summarising reports or automating customer support. Use platforms like Clarifai that provide access to multiple models, data preparation tools and local runners for secure deployment. Implement human‑in‑the‑loop processes and follow ethical guidelines.
Generative AI augments rather than replaces humans. It handles repetitive or data-heavy tasks, freeing people to focus on strategy, creativity and complex decisions.
Use synthetic data where real data is scarce or sensitive. Work with platforms that support local deployment and maintain audit trails. Follow relevant regulations (e.g., GDPR, HIPAA) and consult legal counsel.
Risks include hallucinations (incorrect information), bias propagation, privacy breaches and misuse. Mitigate them by combining RAG with vector databases for factual grounding, performing bias audits and ensuring transparency.
Key modules include Vector Store for context retrieval, Prompt Library, Synthetic Data Generation, Mesh for orchestration and Local Runners for secure deployment
Chip maker Nvidia has officially become the world’s first $5 trillion company.
The company’s value surged past $211 per share this past week, setting yet another record for the chip maker just months after it crossed the $4 trillion mark in July 2025.
The rapid ascent cements Nvidia’s mission-critical position in the AI economy, driven by explosive demand for its processors, the backbone of modern AI. The company now sits ahead of tech giants Apple (at $4 trillion), Microsoft, Alphabet, Amazon, and Meta.
Nvidia’s stock climbed after the company announced a $1 billion purchase of Nokia shares for a partnership to build “AI-native” 5G and 6G networks. It rose further after U.S. President Donald Trump stated he plans to discuss Nvidia’s restricted Blackwell AI chip with Chinese President Xi Jinping.
I discussed Nvidia’s immense growth with SmarterX and Marketing AI Institute founder and CEO Paul Roetzer on Episode 178 of The Artificial Intelligence Show.
Just 18 months ago, Nvidia was valued at under $1 trillion. Today, it’s worth more than Amazon and Meta combined.
To put this staggering growth in perspective, Roetzer offered a stunning statistic.
“On November 30, 2022, the day ChatGPT came out, NVIDIA’s market cap was $422 billion,” Roetzer says.
That single date marks the beginning of the rapid growth of generative AI, with Nvidia being a key driver and beneficiary of it.
Much of the company’s success and its public persona is tied to its CEO, Jensen Huang. Roetzer notes that Huang’s leadership style is as unique as the company’s growth.
“The guy’s awesome,” says Roetzer, noting his ability to be present and focused. “Any interview you ever watch, he is just in the moment.”
In an industry defined by breakneck speed and immense pressure, that kind of focus stands out, Roetzer said, explaining Huang’s grounded approach is a reason to root for the company beyond its meteoric stock price
“I’m happy he’s being successful,” says Roetzer. “And I think he’s the kind of person we want at the forefront of all of this, for the good of society and humanity.”
Companies are getting results with generative AI, according to a new report by The Wharton School of the University of Pennsylvania.
The third-annual report, “Accountable Acceleration: Gen AI Fast-Tracks Into the Enterprise,” tracked corporate AI adoption with a survey of about 800 senior decision-makers. It found that 75 percent of business leaders report a positive return on investment (ROI) from their AI investments. Fewer than 5 percent say returns have been negative.
The data also shows usage is surging: 46 percent of leaders now use generative AI daily (a sharp increase from last year), and 82 percent use it at least weekly. The most common uses are data analysis, document summarization, and editing.
This new data paints a very different picture from other reports that have dominated headlines, namely, a viral MIT report that said 95 percent of generative AI pilots are failing.
To understand why this Wharton data is so different and what it means for business leaders, I turned to SmarterX and Marketing AI Institute founder and CEO Paul Roetzer on Episode 178 of The Artificial Intelligence Show.
Roetzer’s immediate takeaway? The Wharton study, conducted between June and July 2025 with a clear methodology is “legitimate research.”
He contrasts this sharply with the 95 percent failure rate statistic that MIT found, which he thinks was based on flawed methodology.
This new data shows that investment is not only continuing but accelerating. The study found 88 percent of companies anticipate Gen AI budget increases, with 62 percent expecting increases of 10 percent or more.
Roetzer also highlighted one finding: “One third of Gen AI technology budgets are being allocated to internal R&D, which I thought was interesting and an indication that many enterprises are building custom capabilities for the future.”
While adoption and return on investment are high, the Wharton report pinpoints a major barrier: human capital.
The biggest challenges cited by leaders were recruiting talent with advanced generative AI technical skills (49 percent) and providing effective training programs (46 percent). A lack of training resources also emerged as a top barrier for the first time since they’ve been doing the report.
This data highlights a massive opportunity for anyone willing to step up and provide guidance.
Roetzer noted that the people who are already learning AI, such as those actively following the space, are most likely ahead of other people. It’s important they take the lead with their peers and companies.
“You have to pull them along,” Roetzer says. “We all kind of have this responsibility to figure this stuff out and then find ways to help our peers get through this.”
The challenge is that many people are simply afraid or find AI too abstract, Roetzer said, so they never get started. But if they don’t start, they’ll be left behind.
“The future isn’t going to go well for people who don’t learn to embrace this stuff in a responsible way,” he says.
The solution, according to Roetzer, is for internal champions to emerge and help drive the organization forward, translating the abstract into practical applications and overcoming the fear.
“We need those internal champions to help drive this forward,” he said.
Most AI development begins locally. You experiment with model architectures, fine-tune them on small datasets, and iterate until the results look promising. But when it’s time to test the model in a real-world pipeline, things quickly become complicated.
You usually have two choices: upload the model to the cloud even for simple testing, or set up your own API, managing routing, authentication, and security just to run it locally.
Neither approach works well if you’re:
Working on smaller or resource-limited projects
Needing access to local files or private data
Building for edge or on-prem environments where cloud access isn’t practical
Local Runners let you serve AI models, MCP servers, or agents directly from your laptop, workstation, or internal server, securely and seamlessly via a Public API. You don’t need to upload your model or manage any infrastructure. Simply run it locally, and Clarifai takes care of the API handling, routing, and integration.
Once running, the Local Runner establishes a secure connection to Clarifai’s control plane. Any API request sent to your model is routed to your machine, processed locally, and returned to the client. From the outside, it behaves like a Clarifai-hosted model, while all computation occurs on your local hardware.
With Local Runners, you can:
Now that you understand the benefits and capabilities of Local Runners, let’s see how you can run Hugging Face models locally and expose them securely.
The Hugging Face Toolkit in Clarifai CLI enables you to download, configure, and run Hugging Face models locally while exposing them securely through a public API. You can test, integrate, and iterate on models directly from your local environment without managing any external infrastructure.
First, install the Clarifai Package. This also provides the Clarifai CLI:
Next, log in to Clarifai to link your local environment to your account. This allows you to manage and expose your models.
Follow the prompts to enter your User ID and Personal Access Token (PAT). If you need help obtaining these, refer to the documentation.
If you plan to access private Hugging Face models or repositories, generate a token from your Hugging Face account settings and set it as an environment variable:
Finally, install the Hugging Face Hub library to enable model downloads and integration:
With these steps complete, your environment is ready to initialize and run Hugging Face models locally with Clarifai.
Use the Clarifai CLI to initialize and configure any supported Hugging Face model locally with the Toolkit:
By default, this command downloads and sets up the unsloth/Llama-3.2-1B-Instruct model in your current directory.
If you want to use a different model, you can specify it with the --model-name flag and pass the full model name from Hugging Face. For example:
Note: Some models can be very large and require significant memory or GPU resources. Make sure your machine has enough compute capacity to load and run the model locally before initializing it.
Now, once you run the above command, the CLI will scaffold the project for you. The generated directory structure will look like this:
model.py – Contains the logic for loading the model and running predictions.
config.yaml – Holds model metadata, compute resources, and checkpoint configuration.
requirements.txt – Lists the Python dependencies required for your model.
model.pyOnce your project scaffold is ready, the next step is to configure your model’s behavior in model.py. By default, this file includes a class called MyModel that extends ModelClass from Clarifai. Inside this class, you’ll find four main methods ready for use:
load_model() – Loads checkpoints from Hugging Face, initializes the tokenizer, and sets up streaming for real-time output.
predict() – Handles single-prompt inference and returns responses. You can adjust parameters such as max_tokens, temperature, and top_p.
generate() – Streams outputs token by token, useful for live previews.
chat() – Manages multi-turn conversations and returns structured responses.
You can use these methods as-is, or customize them to fit your specific model behavior. The scaffold ensures that all core functionality is already implemented, so you can get started with minimal setup.
config.yamlThe config.yaml file defines model metadata and compute requirements. For Local Runners, most defaults work, but it’s important to understand each section:
checkpoints – Specifies the Hugging Face repository and token for private models.inference_compute_info – Defines compute requirements. For Local Runners, you can typically use defaults. When deploying on dedicated infrastructure, you can customize accelerators, memory, and CPU based on the model requirements.
model – Contains metadata such as app_id, model_id, model_type_id, and user_id. Replace YOUR_USER_ID with your own Clarifai user ID.
Finally, the requirements.txt file lists all Python dependencies required for your model. You can add any additional packages your model needs to run.
Once your model is configured, you can launch it locally using the Clarifai CLI:
This command starts a Local Runner instance on your machine. The CLI automatically handles all necessary setup, so you don’t need to manually configure infrastructure.
After the Local Runner starts, you’ll receive a public Clarifai URL. This URL acts as a secure gateway to your locally running model. Any requests made to this endpoint are routed to your local environment, processed by your model, and returned through the same endpoint.
Once your Hugging Face model is running locally and exposed via the Clarifai Local Runner, you can send inference requests to it from anywhere — using either the OpenAI-compatible endpoint or the Clarifai SDK.
Use the OpenAI client to send a request to your locally running Hugging Face model:
You can also interact directly through the Clarifai SDK, which provides a lightweight interface for inference:
You can also experiment with:
With this setup, your Hugging Face model runs entirely on your local hardware — yet remains accessible via Clarifai’s secure public API.
Local Runners give you full control over where your models run — without sacrificing integration, security, or flexibility.
You can prototype, test, and serve real workloads on your own hardware while still using Clarifai’s platform to route traffic, handle authentication, and scale when needed.
You can try Local Runners for free with the Free Tier, or upgrade to the Developer Plan at $1/month for the first year to connect up to 5 Local Runners with unlimited hours. Read more in the documentation here to get started.
Google Labs has launched a new experimental AI marketing tool called Pomelli, built to automate campaign creation while staying true to your brand’s voice.
The tool works by analyzing a company’s website to understand its identity, tone, and audience. From there, it generates custom marketing campaigns, including headlines, social posts, and ad copy, all designed to sound like they came from the brand itself.
The goal is to make scalable, on-brand content accessible to smaller teams that lack dedicated marketing resources, according to Google Labs. The tool is currently only available in the U.S., Canada, Australia, and New Zealand. Some users have reported hitting usage limits.
How well does it actually work?
To find out, I turned to the expertise of SmarterX and Marketing AI Institute founder and CEO Paul Roetzer during Episode 178 of The Artificial Intelligence Show.
Pomelli’s process starts by generating your “Business DNA.” Users give it a website, and it goes to work analyzing the site to learn its visual aesthetics, tone of voice, and brand values, and even pulls logos and fonts.
This part, Roetzer notes, is impressive. It’s an automated version of a task that would typically take a human at a marketing agency three to five hours.
“It pulls all the images from the website and puts them into a little library that it can then use for creative” work, says Roetzer. “It creates a color palette. It is kind of cool.”
After building the brand profile, the tool asks what campaign you want to run. Roetzer tested it by asking it to “grow our artificial intelligence show podcast audience,” and selected a campaign theme called “Essential AI Insights Weekly.”
This is where the experiment fell apart.
When Roetzer tried to edit the creative assets, the results were less than intelligent. He attempted to swap in a different image of himself standing on stage.
“It cut me off,” he says. “The image it dropped in is my right shoulder and then just a background from the stage. And I was like, ‘OK, this obviously isn’t very intelligent.’ It doesn’t even know to focus on the human.”
This hands-on test led Roetzer to a clear conclusion: “Marketers and creatives, don’t worry. This is not automating your job.”
While the concept is strong, the execution is not there yet. The tool’s failure to handle a simple creative edit was disqualifying in Roetzer’s one test.
“I would not be running a second test,” he says. “It’s not even worth trying to do this again.”
He noted that the tool could improve, but it would require significant fine-tuning from people who actually understand creative work.
“I would say they need to go hire 50 creatives who’ve actually done creative work and do some fine tuning on this model,” says Roetzer. “As it is, it is not there yet.”
DeepSeek models, including DeepSeek‑R1 and DeepSeek‑V3.1, are accessible directly through the Clarifai platform. You can get started without needing a separate DeepSeek API key or endpoint.
https://api.clarifai.com/v2/ext/openai/v1
Authenticate with your Personal Access Token (PAT) and specify the model URL, such as DeepSeek‑R1 or DeepSeek‑V3.1.
Clarifai handles all hosting, scaling, and orchestration, letting you focus purely on building your application and using the model’s reasoning and chat capabilities.
DeepSeek encompasses a range of large language models (LLMs) designed with diverse architectural strategies to optimize performance across various tasks. While some models employ a Mixture-of-Experts (MoE) approach, others utilize dense architectures to balance efficiency and capability.
DeepSeek-R1 is a dense model that integrates reinforcement learning (RL) with knowledge distillation to enhance reasoning capabilities. It employs a standard transformer architecture augmented with Multi-Head Latent Attention (MLA) to improve context handling and reduce memory overhead. This design enables the model to achieve high performance in tasks requiring deep reasoning, such as mathematics and logic.
DeepSeek-V3 adopts a hybrid approach, combining both dense and MoE components. The dense part handles general conversational tasks, while the MoE component activates specialized experts for complex reasoning tasks. This architecture allows the model to efficiently switch between general and specialized modes, optimizing performance across a broad spectrum of applications.
To provide more accessible options, DeepSeek offers distilled versions of its models, such as DeepSeek-R1-Distill-Qwen-7B. These models are smaller in size but retain much of the reasoning and coding capabilities of their larger counterparts. For instance, DeepSeek-R1-Distill-Qwen-7B is based on the Qwen 2.5 architecture and has been fine-tuned with reasoning data generated by DeepSeek-R1, achieving strong performance in mathematical reasoning and general problem-solving tasks.
DeepSeek models can be accessed on Clarifai in three ways: through the Clarifai Playground UI, via the OpenAI-compatible API, or using the Clarifai SDK. Each method provides a different level of control and flexibility, allowing you to experiment, integrate, and deploy models according to your development workflow.
The Playground provides a fast, interactive environment to test prompts and explore model behavior.
You can select any DeepSeek model, including DeepSeek‑R1, DeepSeek‑V3.1, or distilled versions available on the community. You can input prompts, adjust parameters such as temperature and streaming, and immediately see the model responses. The Playground also allows you to compare multiple models side by side to test and evaluate their responses.
Within the Playground itself, you have the option to view the API section, where you can access code snippets in multiple languages, including cURL, Java, JavaScript, Node.js, the OpenAI-compatible API, the Clarifai Python SDK, PHP, and more.
You can select the language you need, copy the snippet, and directly integrate it into your applications. For more details on testing models and using the Playground, see the Clarifai Playground Quickstart
Try it: The Clarifai Playground is the fastest way to test prompts. Navigate to the model page and click “Test in Playground”.
Clarifai provides a drop-in replacement for the OpenAI API, allowing you to use the same Python or TypeScript client libraries you are familiar with while pointing to Clarifai’s OpenAI-compatible endpoint. Once you have your PAT set as an environment variable, you can call any Clarifai-hosted DeepSeek model by specifying the model URL.
Python Example
import os
from openai import OpenAI
client = OpenAI(
base_url=“https://api.clarifai.com/v2/ext/openai/v1”,
api_key=os.environ[“CLARIFAI_PAT”]
)
response = client.chat.completions.create(
model=“https://clarifai.com/deepseek-ai/deepseek-chat/models/DeepSeek-R1”,
messages=[
{“role”: “system”, “content”: “You are a helpful assistant.”},
{“role”: “user”, “content”: “Tell me a three sentence bedtime story about a unicorn.”}
],
max_completion_tokens=100,
temperature=0.7
)
print(response.choices[0].message.content)
TypeScript Example
import OpenAI from “openai”;
const client = new OpenAI({
baseURL: “https://api.clarifai.com/v2/ext/openai/v1”,
apiKey: process.env.CLARIFAI_PAT,
});
const response = await client.chat.completions.create({
model: “https://clarifai.com/deepseek-ai/deepseek-chat/models/DeepSeek-R1”,
messages: [
{ role: “system”, content: “You are a helpful assistant.” },
{ role: “user”, content: “Who are you?” }
],
});
console.log(response.choices?.[0]?.message.content);
Clarifai’s Python SDK simplifies authentication and model calls, allowing you to interact with DeepSeek models using concise Python code. After setting your PAT, you can initialize a model client and make predictions with just a few lines.
import os
from clarifai.client import Model
model = Model(
url=“https://clarifai.com/deepseek-ai/deepseek-chat/models/DeepSeek-V3_1”,
pat=os.environ[“CLARIFAI_PAT”]
)
response = model.predict(
prompt=“What is the future of AI?”,
max_tokens=512,
temperature=0.7,
top_p=0.95,
thinking=“False”
)
print(response)
For modern web applications, the Vercel AI SDK provides a TypeScript toolkit to interact with Clarifai models. It supports the OpenAI-compatible provider, enabling seamless integration.
import { createOpenAICompatible } from “@ai-sdk/openai-compatible”;
import { generateText } from “ai”;
const clarifai = createOpenAICompatible({
baseURL: “https://api.clarifai.com/v2/ext/openai/v1”,
apiKey: process.env.CLARIFAI_PAT,
});
const model = clarifai(“https://clarifai.com/deepseek-ai/deepseek-chat/models/DeepSeek-R1”);
const { text } = await generateText({
model,
messages: [
{ role: “system”, content: “You are a helpful assistant.” },
{ role: “user”, content: “What is photosynthesis?” }
],
});
console.log(text);
This SDK also supports streaming responses, tool calling, and other advanced features.In addition to the above, DeepSeek models can also be accessed via cURL, PHP, Java, and other languages. For a complete list of integration methods, supported providers, and advanced usage examples, refer to the documentation.
DeepSeek models on Clarifai support advanced inference features that make them suitable for production-grade workloads. You can enable streaming for low-latency responses, and tool calling to let the model interact dynamically with external systems or APIs. These capabilities work seamlessly through Clarifai’s OpenAI-compatible API.
Streaming returns model output token by token, improving responsiveness in real-time applications like chat interfaces or dashboards. The example below shows how to stream responses from a DeepSeek model hosted on Clarifai.
import os
from openai import OpenAI
# Initialize the OpenAI-compatible client for Clarifai
client = OpenAI(
base_url=“https://api.clarifai.com/v2/ext/openai/v1”,
api_key=os.environ[“CLARIFAI_PAT”]
)
# Create a chat completion request with streaming enabled
response = client.chat.completions.create(
model=“https://clarifai.com/deepseek-ai/deepseek-chat/models/DeepSeek-V3_1”,
messages=[
{“role”: “system”, “content”: “You are a helpful assistant.”},
{“role”: “user”, “content”: “Explain how transformers work in simple terms.”}
],
max_completion_tokens=150,
temperature=0.7,
stream=True
)
print(“Assistant’s Response:”)
for chunk in response:
if chunk.choices and chunk.choices[0].delta and chunk.choices[0].delta.content is not None:
print(chunk.choices[0].delta.content, end=“”)
print(“\n”)
Streaming helps you render partial responses as they arrive instead of waiting for the entire output, reducing perceived latency.
Tool calling enables a model to invoke external functions during inference, which is especially useful for building AI agents that can interact with APIs, fetch live data, or perform dynamic reasoning. DeepSeek-V3.1 supports tool calling, allowing your agents to make context-aware decisions. Below is an example of defining and using a tool with a DeepSeek model.
import os
from openai import OpenAI
# Initialize the OpenAI-compatible client for Clarifai
client = OpenAI(
base_url=“https://api.clarifai.com/v2/ext/openai/v1”,
api_key=os.environ[“CLARIFAI_PAT”]
)
# Define a simple function the model can call
tools = [
{
“type”: “function”,
“function”: {
“name”: “get_weather”,
“description”: “Returns the current temperature for a given location.”,
“parameters”: {
“type”: “object”,
“properties”: {
“location”: {
“type”: “string”,
“description”: “City and country, for example ‘New York, USA'”
}
},
“required”: [“location”],
“additionalProperties”: False
}
}
}
]
# Create a chat completion request with tool-calling enabled
response = client.chat.completions.create(
model=“https://clarifai.com/deepseek-ai/deepseek-chat/models/DeepSeek-V3_1”,
messages=[
{“role”: “user”, “content”: “What is the weather like in New York today?”}
],
tools=tools,
tool_choice=‘auto’
)
# Print the tool call proposed by the model
tool_calls = response.choices[0].message.tool_calls
print(“Tool calls:”, tool_calls)
For more advanced inference patterns, including multi-turn reasoning, structured output generation, and extended examples of streaming and tool calling, refer to the documentation
Clarifai hosts multiple DeepSeek variants. Choosing the right one depends on your task:
At the time of writing, DeepSeek‑OCR has just been announced and is not yet available on Clarifai. Keep an eye on Clarifai’s model catalog for updates.
Q1: Do I need a DeepSeek API key?
No. When using Clarifai, you only need a Clarifai Personal Access Token. Do not use or expose the DeepSeek API key unless you are calling DeepSeek directly (which this guide does not cover).
Q2: How do I switch between models in code?
Change the model value to the Clarifai model ID, such as openai/deepseek-ai/deepseek-chat/models/DeepSeek-R1 for R1 or openai/deepseek-ai/deepseek-chat/models/DeepSeek-V3.1 for V3.1.
Q3: What parameters can I tweak?
You can adjust temperature, top_p and max_tokens to control randomness, sampling breadth and output length. For streaming responses, set stream=True. Tool calling requires defining a tool schema.
Q4: Are there rate limits?
Clarifai enforces soft rate limits per PAT. Implement exponential backoff and avoid retrying 4XX errors. For high throughput, contact Clarifai to increase quotas.
Q5: Is my data secure?
Clarifai processes requests in secure environments and complies with major data‑protection standards. Store your PAT securely and avoid including sensitive data in prompts unless necessary.
Q6: Can I fine‑tune DeepSeek models?
DeepSeek models are MIT‑licensed. Clarifai plans to offer private hosting and fine‑tuning for enterprise customers in the near future. Until then, you can download and fine‑tune the open‑source models on your own infrastructure.
You now have a fast, standard way to integrate DeepSeek models, including R1, V3.1, and distilled variants, into your applications. Clarifai handles all infrastructure, scaling, and orchestration. No separate DeepSeek key or complex setup is needed. Try the models today through the Clarifai Playground or API and integrate them into your applications.
A new paper from the Center for AI Safety and Scale AI has introduced the Remote Labor Index (RLI), the first benchmark designed to measure how well AI agents can perform paid, remote jobs.
The RLI benchmark includes real-world projects from freelance platforms, spanning complex fields such as game development, architecture, data analysis, and video production. These aren’t simple tasks: The projects represented over 6,000 hours of human work valued at more than $140,000.
The results? Current AI agents performed poorly.
Manus, the top-performing agent, could only automate 2.5 percent of the work. Other top models, such as Grok 4 and Sonnet 4.5, managed just 2.1 percent, while GPT-5 hit 1.7 percent and Gemini 2.5 Pro came in under 1 percent. The researchers noted failures stemmed from incomplete deliverables, broken files, and low-quality work that wouldn’t meet professional standards.
While these low numbers might seem reassuring to human workers, they don’t tell the whole story. To understand what these findings really mean for the future of AI in the workforce, I discussed them with SmarterX and Marketing AI Institute founder and CEO Paul Roetzer on Episode 178 of The Artificial Intelligence Show.
Roetzer wasn’t surprised by the low automation rates, noting that the benchmark tests general agents that aren’t specifically trained for these complex jobs.
The real and much faster progress is happening with specialized agents. He points to examples including OpenAI reportedly hiring Goldman Sachs bankers to train models to do the job of an investment banker.
“My guess is OpenAI’s is way further along than 2.5 percent for that specific thing,” he says.
This highlights a crucial distinction in how we should think about AI’s capabilities. The RLI provides a valuable baseline for general models, but the true economic impact will likely come from models intensely focused on a specific job.
Roetzer explains this using a simple framework: tasks, projects, and jobs.
Right now, AI is very good at the task level, which includes the small, discrete activities that make up a larger project.
“It’s good at the tasks,” he says. “It’s not good at doing the full thing.”
An agent can’t replace a CEO, for example, but it might help with 25 different tasks that a CEO does every month. Humans, however, are still essential for setting goals, planning, connecting data sources, integrating tools, and, most importantly, overseeing and verifying the AI output.
The key metric to watch, according to Roetzer, is how long an agent can work without a human needing to intervene, a concept he calls “actions per disengagement,” similar to how Tesla measures self-driving.
We haven’t yet reached what he calls the “economic Turing test,” where the economic labor of AI is indistinguishable from that of a human.
“Is it to the point where I would hire an agent or a symphony of agents instead of a human?” he asks. “In every instance I can think of, the answer is still no.”
However, agents are getting better, more autonomous, and more reliable within specific jobs slowly but surely. And even augmentation of people with AI agents may lead to a reduction in the number of people needed, says Roetzer.
“As the agents get more autonomous, as they get more reliable, as more companies understand how to build and integrate them into workflows, you don’t need as many people doing the work that you previously did.”
Artificial intelligence isn’t just about what models can do—it’s about where they run and how they deliver insights. In the age of connected devices, Edge AI and Cloud AI represent two powerful paradigms for deploying AI workloads, and enterprises are increasingly blending them to optimize latency, privacy, and scale. This guide explores the differences between edge and cloud, examines their benefits and trade‑offs, and provides practical guidance on choosing the right architecture. Along the way, we weave in expert insights, market data, and Clarifai’s compute orchestration solutions to help you make informed decisions.
Edge AI refers to running AI models locally on devices or near the data source—for example, a smart camera performing object detection or a drone making navigation decisions without sending data to a remote server. Edge devices process data in real time, often using specialized chips or lightweight neural networks, and only send relevant insights back to the cloud when necessary. This eliminates dependency on internet connectivity and drastically reduces latency.
At its core, edge AI moves computation from centralized data centers to the “edge” of the network. Here’s why companies choose edge deployments:
These benefits explain why 97 % of CIOs have already deployed or plan to deploy edge AI, according to a recent industry survey.
Imagine a smart camera in a factory that can instantly detect a defective product on the conveyor belt and stop the line. If it relied on a remote server, network delays could result in wasted materials. Edge AI ensures the decision happens in microseconds, preventing expensive product defects.
Cloud AI refers to running AI workloads on centralized servers hosted by cloud providers. Data is sent to these servers, where high‑end GPUs or TPUs train and run models. The results are then returned via the network. Cloud AI excels at large‑scale training and inference, offering elastic compute resources and easier maintenance.
Key characteristics of cloud AI include:
A retailer might run a massive recommendation engine in the cloud, training it on millions of purchase histories. Each store then downloads a lightweight model optimized for its local inventory, while the central model continues learning from aggregated data and pushing improvements back to the edge.
Edge and cloud AI differ primarily in where data is processed and how quickly insights are delivered. The edge runs models on local devices for low latency and privacy, while the cloud centralizes computation for scalability and collaborative training. A hybrid architecture combines both to optimize performance.
|
Feature |
Edge AI |
Cloud AI |
|
Processing location |
On-device or near‑device (gateways, sensors) |
Centralized data centers |
|
Latency |
Milliseconds; ideal for real‑time control |
Seconds; dependent on network |
|
Data privacy |
High—data stays local |
Lower—data transmitted to the cloud |
|
Bandwidth & connectivity |
Minimal; can operate offline |
Requires stable internet |
|
Scalability |
Limited by device resources |
Virtually unlimited compute and storage |
|
Cost model |
Upfront hardware cost; lower operational expenses |
Pay‑as‑you‑go but can become expensive over time |
|
Use cases |
Real‑time control, IoT, AR/VR, autonomous vehicles |
Model training, large-scale analytics, generative AI |
In a smart city, traffic cameras use edge AI to count vehicles and detect incidents. Aggregated counts are sent to a cloud AI platform that uses historical data and weather forecasts to optimize traffic lights across the city. This hybrid approach ensures both real‑time response and long‑term planning.
Edge AI delivers ultra‑low latency, enhanced privacy, reduced network dependency and cost savings. It’s ideal for scenarios where rapid decision‑making, data sovereignty or unreliable connectivity are critical..
An agritech startup deploys edge AI sensors across remote farms. Each sensor analyses soil moisture and weather conditions in real time. When a pump needs activation, the device triggers irrigation locally without waiting for central approval, ensuring crops aren’t stressed during network downtime.
Cloud AI excels at scalability, high compute performance, centralized management and rapid innovation. It’s ideal for training large models, global analytics and orchestrating updates across distributed systems.
A biotech firm uses the cloud to train a protein‑folding model on petabytes of genomic data. The resulting model helps researchers understand complex disease mechanisms. Because the data is centralized, scientists across the globe collaborate seamlessly on the same datasets without shipping data to local clusters.
While edge and cloud AI offer significant advantages, both have limitations. Edge AI faces limited compute and battery constraints, while cloud AI contends with latency, privacy concerns and escalating costs. Navigating these trade‑offs is essential for enterprise success.
A hospital deploys AI for patient monitoring. On‑premises devices detect anomalies like irregular heartbeats in real time, while cloud AI analyzes aggregated data to refine predictive models. This hybrid setup balances privacy and real‑time intervention but requires careful coordination to keep models synchronized and ensure regulatory compliance.
The choice depends on latency requirements, data sensitivity, connectivity, cost constraints and regulatory context. In many cases, the optimal solution is a hybrid architecture that uses the cloud for training and coordination and the edge for real‑time inference.
Smart retail systems use edge cameras to track customer foot traffic and shelf interactions. The store’s cloud platform aggregates patterns across locations, predicts product demand and pushes restocking recommendations back to individual stores. This synergy improves operational efficiency and customer experience.
Emerging trends include edge LLMs, tiny models, 5G, specialized chips, quantum computing and increasing regulatory scrutiny. These innovations will broaden AI adoption while challenging companies to manage complexity.
Imagine a future where wearables run personalized LLMs that coach users through their daily tasks, while the cloud trains new behavioral patterns from anonymized data. Such a setup would blend personal privacy with collective intelligence.
AI is transforming industries from manufacturing and healthcare to retail and transportation. Enterprises are adopting edge, cloud and hybrid solutions to enhance efficiency, safety and customer experiences.
At a distribution center, robots equipped with edge AI navigate aisles, pick orders and avoid collisions. Cloud dashboards track throughput and suggest improvements, while federated learning ensures each robot benefits from the collective experience without sharing raw data.
Clarifai offers compute orchestration, model inference and local runners that simplify deploying AI models across cloud, on‑premises and edge environments. These tools help optimize costs, ensure security and improve scalability.
Clarifai’s compute orchestration provides a unified control plane for deploying any model on any hardware—cloud, on‑prem or air‑gapped environments. It uses GPU fractioning, autoscaling and dynamic scheduling to reduce compute requirements by up to 90 % and handle 1.6 million inference requests per second. By avoiding vendor lock‑in, enterprises can route workloads to the most cost‑effective or compliant infrastructure.
With Clarifai’s inference platform, organizations can make prediction calls efficiently across clusters and node pools. Compute resources scale automatically based on demand, ensuring consistent performance. Customers control deployment endpoints, which means they decide whether inference happens in the cloud or on edge hardware.
Clarifai’s local runners allow you to run and test models on local hardware while exposing them via Clarifai’s API, ensuring secure development and offline processing. Local runners seamlessly integrate with compute orchestration, making it easy to deploy the same model on a laptop, a private server or an edge device with no code changes.
An insurance company deploys Clarifai’s compute orchestration to run vehicle damage assessment models. In remote regions, local runners analyze photos on a claims agent’s tablet, while in urban areas, the same model runs on cloud clusters for rapid batch processing. This setup reduces costs and speeds up claims approvals.
Edge AI processes data locally, so raw data doesn’t leave the device. Only aggregated insights or model updates are transmitted to the cloud. This reduces exposure to breaches and supports compliance with regulations like HIPAA and the EU AI Act.
Edge AI requires upfront investment in specialized hardware, but it reduces long‑term cloud costs. OTAVA reports cost savings of 30–40 % when offloading inference to the edge. Cloud AI charges based on usage; for heavy workloads, costs can accumulate quickly.
Industries with real‑time or sensitive applications—manufacturing, healthcare, autonomous vehicles, retail and agriculture—benefit greatly. These sectors gain from low latency, privacy and offline capabilities.
Hybrid AI refers to combining cloud and edge AI. Models are trained in the cloud, deployed at the edge and continuously improved through feedback loops. This approach maximizes performance while managing cost and compliance.
Clarifai’s compute orchestration, local runners and model inference provide an end‑to‑end platform for deploying AI across any environment. These tools optimize compute usage, ensure security and enable enterprises to harness both edge and cloud AI benefits.
The debate between edge and cloud AI isn’t a matter of one replacing the other—it’s about finding the right balance. Edge AI empowers devices with lightning‑fast responses and privacy‑preserving intelligence, while cloud AI supplies the muscle for training, large‑scale analytics and global collaboration. Hybrid architectures that blend edge and cloud will define the next decade of AI innovation, enabling enterprises to deliver immersive experiences, optimize operations and meet regulatory demands. As you embark on this journey, leverage platforms like Clarifai’s compute orchestration and local runners to simplify deployment, control costs and accelerate time to value. Stay informed about emerging trends, invest in skill development, and design AI systems that respect users, regulators and our planet.
Federal Reserve Chair Jerome Powell just delivered a sobering warning: The US labor market is slowing down, and AI is a key reason why. Continue reading “AI Is Impacting the Job Market”
