Why has talk of AI slowed down?

To start, yes AI has reached a transformative stage and, if you're in an industry where is was big during Chap GPT tsunami you probably should be trying to find out how to incorporate it into your daily work.

Beyond that though, why has talk of AI slowed down so drastically?

It is impossible to know all of the reasons, but I think that the biggest reason is actually a pretty simple one; it was over-hyped.

This is not the same as saying "it is all hype". There was definitely some validity to the claims. Hence my opening statement to adopt it if you can.

What has happened since is that a lot of people and companies have tried to make entire products or businesses or whatever based around AI and quickly run into a few problems:

• For the lay-person AI can only generate a "complete" solution for a basic tier version of that solution.
• Competitive solutions require professionals of greater caliber than the AI itself to validate and adjust output.

For the first point, a "complete" solution refers to, more or less what it sounds like. A fully accurate, functional and deliverable unit of work. As a programmer this would mean that once I'm done feeding in my original prompt, I get code for a complete solution (front and back end), the code is syntactically correct and the code also does, logically, what it was asked to do.

The SECOND you need to refine your prompts or need to do more than the most basic work to get a solution compiled and running you've "broken the seal" and need to have actually developer skills and training.

And this is where current AI falls apart. It is trained on a broad set of data. If you ask it to do something it has been trained on already, then it will probably do pretty good. However, if you're just asking it to do something which has already been done, then you're not bringing any competitive advantage. It is also likely to perform on those well known tasks in a more average capacity. It is unlikely to meet or exceed the quality of the best examples from the training set. 

Once you start asking it to do something novel, it is going to get worse. It is most likely to pluck a combination of things it already knows and patch them together rather than attempt to create anything particularly novel itself. And this is because delivering quality, novel ideas is a process which is orders of magnitude more challenging than rephrasing and rehashing something your already know.

When Chat GPT first exploded in popularity a few things happened. Firstly, professionals (especially developers) who had worked with AI tools before were blown away by the advances. Chat GPT 3.5 and 4 are so far ahead of older models that it was quite easy to get carried away. As time has gone on, they have tempered themselves as they have hit more and more "hallucinations" or just flat out failures. 

Another thing which fizzled out is the profit from the tool. And this is for much the same reason. Early adopters were able to turn out some solutions quickly into less saturated markets. However, if YOU built a particular app or website using JUST AI, then so can anyone else with access. Once the market becomes saturated, then the average quality stagnates and people need to know more about the underlying field to keep ahead of the pack.

Most amusing to me though is the knock-on effect. Now that Chat GPT is here, there is EVEN more data available. But, more and more of that data is generated by AI itself and is of a lesser quality. The emerging problem for AI will become figuring out how to filter out AI generated data from their data sets so that it doesn't pollute the pool and create natural pockets of data. Because in those cases the AI will actually become much less intelligent over time. With increasingly more of the content striking the average, the actually good data will become increasingly under emphasized.

And my thought is that these are the reasons for the decline in AI reporting lately. The shine has worn off and we're much more acutely aware of the deficiencies combined with a race to the bottom on producing products and services based solely on AI.