These are the Real Trends in Generative AI

What’s real and what’s noise in Generative AI? Beyond the buzzwords and LinkedIn hype.

I've researched 2,651 AI projects from a major freelance platform for industry, problem, and AI solution requested.

Here's what I've found:

Background: I wanted to know what's hot in the "real" Generative AI world at the moment.

Away from the buzz all the AI cowboys are marking on LinkedIn.

So I went to look at the raw market data.

The points in the image represent 2,651 project postings scraped from the respective platform.

What you see:

• Each small dot represents a project and its requested solution, e.g. “Image annotation service” or “AI agents development” (labels not shown here for clarity)

• The closer two dots to each other, the closer the solutions, e.g. “AI automation solutions” is closer to “Automation workflow development” than to “Voice cloning tool”

• Groups of similar solutions are grouped in their respective “clouds” and named by their respective topic, e.g. “Web Application Development Services”.

The data represents the entire category of AI projects of said freelance platform. Hence not all topics are about Generative AI, such as Machine Learning model developments or predictive systems.

The top ten solution clusters are:

1. Generative AI and Computer Vision Solutions: From image annotation to object detection systems, this cluster is all about enhancing visual data interpretation or generating it. Industries like healthcare, security, and entertainment are rapidly adopting these solutions for efficiency and innovation.

2. Chatbot Development and Integration: Chatbots are evolving beyond simple Q&A to more personalized, conversational AI agents. They're not just for customer service anymore; think mental wellness apps and even storytelling bots.

3. AI Agent Development: We're talking about custom AI agents that can automate complex tasks across various domains. It's like having a team of AI-powered assistants at your disposal.

4. AI-Driven Trading and Financial Automation: The financial sector is buzzing with AI-driven trading bots and algorithmic systems. These tools are revolutionizing how trading is done, offering speed and precision that human traders can't match. Note that most of this is “old-school” machine learning, but AI agents are catching up fast.

5. AI Automation Solutions: From scheduling to process optimization, AI is streamlining workflows across industries. It's all about doing more with less effort and time.

6. Web Application Development Services: From web app development to mobile solutions, these services are crucial for integrating Generative AI into our day-to-day lives.

7. Language Model Development and Evaluation Services: Developing and fine-tuning language models for applications like translations and custom NLP solutions, enhancing AI's language capabilities.

8. AI Integration and Development Services: Custom AI solutions and integration strategies are helping businesses incorporate AI into their operations effectively.

9. Data Services and Solutions: From data quality review to AI-driven SQL optimization, these services provide the backbone for data-driven decision-making.

10. Voice Technology and Audio Processing Services: Innovations in voice cloning, speech-to-text, and music production are transforming how we interact with audio technology.

What’s interesting to observe (and not surprising at all!) is that the majority of the top ten solutions are about Generative AI, putting other classical topics such as “Machine Learning Model Development” or “Deep Learning Model Optimization” into the backseat.

Whether this is temporary or not, only time will tell. But what’s sure is that for many tasks Generative AI is much more capable than traditional predictive models from the past.

The big winners here are “visual” AI, chatbots, AI agents, and Gen AI automation and integration.

That means that the market has a high demand for what these technologies are best at: generating content (text and visual), human-like decision-making through agents, and automating manual workflows through integration.

Again, this is a rather small part of the market. But given this is a major freelancer platform and the straightforwardness of the data (no guesswork or CEO interviews), it should be a pretty accurate representation of what’s going on out there.

How did I prepare this data?

The project postings were very fuzzy, in the sense that the platform didn’t include separate fields on the important data points such as industry and solution. Each project was basically a wall of text.

So here's what I did:

0. Scraped the platform for all its 2,651 jobs from its AI section

1. Uploaded it into Google Sheets

2. Created a CustomGPT specifically instructed to extract the industry and solution from the project postings.

3. Applied the CustomGPT right in Google Sheets via the "GPT for Sheets" extension → Created new columns with the extracted data

4. Packed the extracted info (market & solution) into Python

5. Calculated text "embeddings" (aka number representations) for each entry

6. Grouped (clustered) the points based on the embeddings

7. Calculated the statistics

Note: ChatGPT can't do steps 4-6 out-of-the box. Clustering is an "exact" operation – aka crunching numbers, that are better tackled by specialized software.

GPT 4o-mini is a wonder child! It's ultra cheap, fast, and effective, but ONLY when prompted correctly.

So here are some prompting decisions I took when extracting the data:

• To be cost-effective on 2,651 job postings I used 4o-mini instead of o1 or Deepseek, at a marginal decrease in quality

• Prompted it to act like its bigger brothers o1 and o3 via the "chain-of-though" (CoT) method

• Showed it a few "thinking" examples via sample projects from the platform that I processed manually

You can extract structured information with Generative AI at scale from any kind of data: legal documents, marketing materials, competitor ads, patient data, ecommerce inventory… you name it.

If this sparked your curiosity and you want to know more about data extraction, reply to this message, or schedule a call.

Cheers,
Robert