Artificial Intelligence – The Asymptote pt. II

There is no GPT-5.

• OpenAI has disclosed its medium-term roadmap and in the process, effectively admitted that scaling is coming to an end supporting RFM’s long-held view that superintelligent machines are as far away as ever meaning that the real returns from AI will be made from focusing on what AI is good at rather than chasing a pipe dream.
• Sam Altman has released OpenAI’s current roadmap (see here) with the next version of GPT being named 4.5 rather than 5 in a sign that the unreleased model has not made enough of an advance in performance to be considered next generation.
• GPT-5 will be “a system that integrates a lot of our technology, including o3” which does not in any way sound like a shiny new model that will take OpenAI’s performance one step ahead of everyone else.
• This means that the company intends to consolidate all of the models that it currently has under an umbrella piece of code which will be able to allocate requests to different models depending on the nature of the request and the service tier that the user is paying for.
• Hence, GPT-5 is not a new model but a user interface that combines all of the existing technology into something more useful to the user rather than taking performance to the next level.
• To me, this is an admission that the scaling laws that have observed improvements in model performance as parameters, data and compute are increased (see here) have stopped working.
• There have been signs of this for many months (see here).
  • First, Incremental: which is the fact that models have been improving only incrementally with each generation despite huge increases in resources.
  • Second, clustering: which is the fact that all of the leading models perform pretty much to a similar level.
• Both of these are clear evidence of slowing improvements meaning that the current techniques are hitting a wall beyond which they cannot progress.
• This can only lead to price competition and a correction when everyone realises that the business models assuming continued improvements in machine intelligence are greatly overstated.
• Generative AI has two superpowers which are the ability to use natural language as a man-machine interface and the ability to automatically ingest, store, cross reference and retrieve unstructured data in multiple formats.
• These superpowers will deliver substantial benefits and productivity enhancements allowing plenty of new businesses to flourish but they will not deliver superintelligent machines (AGI).
• This opens the way for foundation models to become the operating systems of AI and currently, there are far more in existence that the market is likely to support.
• Over the next few years, I expect that developers will migrate to creating their services using a foundation model as their starting point and using the tools and services rather than writing directly to the silicon.
• This is what RFM refers to as AI Ecosystem 2.0 and this is likely to be where the real battle for leadership in the AI industry is likely to be fought in the coming years.
• This is the reality of where the AI industry is going, and I think that the only thing that can prevent a correction is AGI but there is no sign of this anywhere on the horizon.
• This will trigger a shakeout where Open AI is likely to be acquired by Microsoft, Anthropic by Amazon as well as many others that merge or simply go out of business.
• Nvidia is one of the ones that will suffer the least as it is the only one that is making real money from generative AI right now and even with a correction in demand for its processors, it will suffer much less than those offering generative AI services for $20 per month.
• Hence, if I were forced into direct investment in the generative AI sector, I would choose Nvidia, but I prefer to look more laterally at inference at the edge or nuclear power to run all of these new data centres that are springing up.
• I have positions in both of these themes.