The AI Winter: some possible scenarios

Here are some possible scenarios for the ‘AI winter’:

1. GPUs are regulated. Or an AI mortarium. This is the solution Eliezer Yudkowsky advocates. The likelihood of this happening is really slim. Good luck getting Congress to ever agree to such a thing, given that four years later TikTok has still not been banned despite a stronger case for TikTok being a threat to national security compared to AI. Same for cryptocurrencies despite being used for fraud and cybercrime. And good luck getting China to ever go along with an AI mortarium unless it voluntarily agrees to it.

2. Major economic crisis/recession, like 2007-2009.

3. The third and imho the most likely outcome is that users and the media simply lose interest. Subsequent iterations of GPT fail to live up to hype.

There was huge surge of interest in Chat GPT after it was released to the pubic in Nov. 2022:

Remember all the hype in May/June 2022 about Dall-E? So much for that:

I remember the dire pronouncements at the time about how Dall-E would render artists unemployable and cause a crisis in the arts by flooding the market with AI-generated knockoffs.

What happened? I am guessing the media and users lost interest. There was a 6 month gap between Dall-E and the public unveiling of Chat GPT, so it’s not like the latter stole the spotlight former. Admittedly, Chat GPT is a much bigger, more comprehensive product than Dall-E, but the pattern seems the same.

4. The overuse of paywalls and the degradation of free functionality of Open AI products can cause users to lose interest, hence less virality on social media. Wolfram Alpha debuted in 2009 to similar media fanfare for its AI capabilities. However, it stopped being as well maintained, put most of its functionality behind a paywall, it’s very buggy , and hardly anyone talks about it or cares about it anymore despite all the attention it got earlier.

Similar to Wolfram Alpha, students are using Chat GPT to circumvent school work, but for writing assignments instead of math assignments. Like Wolfram Alpha, Open AI offers two tiers: a free service and a subscription, which costs $20/month. The free service however puts stress on the servers, and given that students are either incapable or unwilling to spend money on subscription plans, imposes a significant cost on Open AI. This is why the free version of Wolfram Alpha became much less useful and heavily metered, with much of the original free functionality put behind various subscription plans.

5. VCs find a new fad. AI is the latest bubble for 2022-2023 after the crypto/DEFI and ‘WFH’ bubbles imploded in 2021-2022. VCs are trying to create media hype surrounding Open AI in anticipation of an IPO , so they can sell inflated shares on retail suckers/investors. It’s the same pattern as always. In Jan. 2021, Open AI raised capital from four VC firms, including Andreessen Horowitz and Sequoia Capital, which are among the biggest players in VC. Now they are looking to ‘exit’ , which means hopefully an IPO.

Overall, based on my 15+ years of writing about and following and investing in consumer tech, pundits either vastly overestimate or underestimate the influence of technologies. Same for market valuations. Facebook/Meta has been underestimated since its founding: “It’s just another Myspace…it will be a fad…teens/kids will grow up and lose interest in it,” yet over a decade since its IPO it’s worth over half a trillion dollars and earns tens of billions of dollars in annual profits. As for teens losing interest, the experts failed to anticipate that Facebook would capture the highly lucrative ‘boomer’ market, and that teens would move on to Instagram, which is also owned by Meta and has very lucrative, high-paying mobile ads. Or the iPhone: “Blackberry is better…saturated market,” yet 15 years since the debut of the iPhone, Apple is worth over $2 trillion. Or Tesla: “Too expensive, charging stations unfeasible, does not work well in cold weather, limited distance, etc.”

In 2017-2018 ‘experts’ were giving 6-7 figure price forecasts for Bitcoin for 2020+, and here we are in 2023 and Bitcoin is still at $28,000, well below even the most conservative of bullish forecasts and only up 50% since late 2017. In terms of adoption, Bitcoin is still mostly a niche and is nowhere close to replacing or even disrupting the financial system, but as shown by the collapse of Silicon Valley Bank and Signature Bank, seems highly dependent on it and just another highly correlated risky asset. My own take/guess is that AI falls into the same category as Bitcoin, as something in which pundits are expecting too much change, either good or bad.