Topic focused on AI Industry Developments
1 Like
Just a general observation: The release of Gemini 3 showed how weak moats are for LLMs. As soon as a new LLM is better people start to move as they want the best underlying AI.
This mechanic might lead to a slowdown of the AI race as investors like Softbank might realize that a company like OpenAi is not a safe bet and potentially not worth such gigantic valuations.
If this happens valuation of Nvidia could come under pressure. For Meta it might be a positive because there could be at one point less capex in chasing AI supremacy.
If AI continues to evolve rapidly from here and AGI will become possible or people get replaced in large numbers by AI the story could obviously be completely different. But just based on current tech and capabilities the non stickiness of costumers could be a very large problem. How do you convince investors to invest billions of dollars if in the end all your Capex becomes greatly diminished or partially worthless if another company builds a superior model and people switch?
1 Like
I mostly agree with this, my understanding of the LLM chatbot user behaviour remains limited. I would like to document the fact for example that my main chat remains ChatGPT, although I am well aware that for many cases the superior LLMs are elsewhere. The fact that I have my libraries/folders, my pre-written general context, the memory, and I also like the environment…
Nonetheless, although ChatGPT remains my main product, I still make use, daily, of Grok, Perplexity (to access Gemini), and Claude (however this does not count as I am referring to the code plug-in, interestingly for coding this is the only thing I use, however I am a light user).
Think it will be handy to have our current usage behaviour documented, as the future of AI adoption is everyone else becoming superusers like us. When you speak to the average person in the street they still barely use these products, although making use of these products is a no brainer. However, as with dotcom, although technology’s pace is rapid, real user adoption remains human and slow.
Good points. I am not using GPT instructions/general context (need to start) or folders so I cannot compare those aspects with Gemini but I found recently that Gemini often provided me with more concise or better answers than GPT so I started to shift for a lot of queries. (But I keep experimenting and comparing)
I think once Ai gets sophisticated enough to remember all your previous queries then „knowing the user well“ will become a very strong moat.
At the moment I don’t see a lot of differences in this aspect and for example GPT does not answer questions better just because I have used it longer. (again just based on plain use & with no prewritten instructions)