Artificial Intelligence – Lifecycle of a bubble pt. II

Bubble valuations but real use cases. 

• As more and more models are released by companies, start-ups and open source, the harder they will have to work to attract users meaning falling prices which is what I think will burst the bubble.
• With valuations in nosebleed territory and governments falling over themselves to be involved in AI, there is very little doubt that we are in a bubble.
• However, as always, two things can be true at once which is that while AI equity (both public and private) is almost universally overvalued, there are significant improvements that generative AI can bring to both the Digital Life of consumers and to the enterprise.
• The reason why we are in a bubble is because the explosion of ChatGPT onto global consciousness has led many to think that the day when the machines are more intelligent than humans is now much closer than it was 3 or 4 years ago.
• This has caused massive excitement and an inflow of capital into a sector where there is not much equity which in turn led to a run-up in equity price that no reality will ever be able to justify.
• In the spring of 2023 a 1-month-old French company called Mistral AI with no product, no plan, no R&D and no IP raised €105m at a valuation of €240m.
• Just recently, Chinese startup 01.ai which is just 8 months old, raised money in a round led by Alibaba Cloud at a valuation of over $1bn although it does have a 34bn parameter model that objectively performs quite well.
• However, there is no real mention of how the company will actually make revenue let alone profit meaning that there is only pie in the sky to rely on.
• Activity is frenetic with Elon Musk releasing a new chatbot called Grok which is trained on Twitter data and there are now over 300,000 models on Hugging Face which is similar to Github but for AI models.
• Furthermore, with the release of Meta’s foundation model LlaMa (and now LlaMa 2) and some clever tricks, the open-source community is now able to fine-tune models and services on powerful laptops meaning that there are thousands of models available for free.
• The combination of start-ups, companies and open source means that the number of available AI services is increasing at a breakneck pace especially with all the money being pumped into the sector.
• The current mantra (and what everyone seems to be raising money on) is that $20 / user/month is the rate that everyone will be able to earn and at that price, less than 1m subscribers provide a very healthy business by RFM’s calculations.
• However, with so many new models and many available for free, I suspect that there will be a lot of price erosion meaning that the real price is probably something like $20 / user/year or maybe even lower.
• This means that in 12 months when everyone needs to raise more money, they will have badly missed their targets and the long-term outlook will be lower.
• This is what I think pricks the valuation bubble and causes expectations and valuations to reset to reality.
• The reality is that there are real use cases and benefits to be had from generative AI but these machines can’t think and they are not about to put everyone out of work despite what Mr Musk says.
• It was also Mr Musk who was confidently predicting that autonomous vehicles would be in the market and working by 2019 and it was the failure of autonomous driving to generate revenue on schedule that caused the correction in valuation and massive consolidation.
• I suspect that generative AI will have a similar trajectory with the difference that it is able to generate good revenues and profits now it is just that it will generate much less of both than anyone thinks.
• Furthermore, there is no empirical evidence that suggests that these machines can reason and they continue to hallucinate and make obvious mistakes at a frightening rate.
• Consequently, those looking for the dawn of the super-intelligent AI age are going to have to wait a very long time.
• Those looking to hire AI engineers or buy assets may have much more to choose from and at much better prices in 12 to 18 months.