Seven quick thoughts on AI
Will AI take over our personal and professional lives? I don't think it's a sure bet. A few quick thoughts on where I stand.
- No one knows what AI will bring - everyone is guessing (including me). So don't get too hung up on what others have to say.
- AI takeover is not a certainty. I'm not talking about whether we will reach AGI, or the many ethical concerns about AI as a technology, but cold hard economics. None of the major US-based AI labs are close to making a profit as standalone businesses, and the capital investment required in building infrastructure to support AI is dizzying. OpenAI's recent pivot to enterprise use cases highlights the fact that none of the AI labs have a compelling customer offering now.
- Only a small subset of OpenAI's 900m+ active monthly users are paid subscribers, and many people are adamantly rejecting any AI use. People who use AI to answer random questions are unlikely to pay for it (especially when there are many free options). Even those who have more involved use cases, e.g. vibe coding or companionship - unless it's revenue generating, or it delivers real benefits to the end user (benefit, of course, is subjective), won't be paying $US20+ a month for the privilege. If AI doesn't reach the ubiquity of the internet or mobile phones, it's hard to see the AI labs generating the kind of revenue that will make them sustainable.
- Maybe businesses will be the saviours of the AI labs. Certainly, they have more cash to invest. So far, a lot of businesses are in the FOMO stage - invest in AI now even if the returns are not there (yet?). For businesses to continue investment in AI, the benefits will need to be transformational. What does transformation look like? Of course, companies can use AI to open new markets and develop new businesses, but for many it is likely to be automation followed by massive staff cuts. If it happens, some businesses will help their staff manage the tectonic shifts in their industries, but many others will just lay off staff and see the human cost as a societal/governmental problem (I'll have more to say about this in the future).
- Back to point 2 above. It is entirely plausible that AI labs will collapse because their revenue never reaches the heights required and their investors bail out. The following example is at a much smaller scale, but think about 3D cinema. There were a few successes at the start, and many pundits were predicting the death of 2D cinema. The technology involved was certainly dazzling, but 3D movies were costly to make. Most people saw them as gimmicks, and the demand for them dried up eventually.
- If AI as a technology becomes entrenched, I think the most likely scenario is that a small group of smart/adventurous/ruthless people will earn an outsized return from AI. Underneath them, a cohort will use AI effectively and generate some benefits. The majority will use AI in shallower ways and not get anywhere, and there will be those who are left behind. The ultra-rich will probably become even richer. Broader ramifications on society are harder to predict. If we turn to history, the onset of the industrial revolution created massive wealth but also myriad societal problems that took generations to address, and the effects of hollowing out blue collar jobs in many western societies in the last few decades are still playing out.
- I'm giving us humans a fair chance of handling whatever AI may bring and coming out victorious. We may put AI back inside Pandora's Box; it may turn out to be useful but not transformational; or it may indeed turn the world upside down, but we will navigate the turbulence and make something out of it. I believe in us.
I just read this very considerate piece from Ezra Klein on how AI can affect our sense of self.
A.I. sycophancy — the tendency of these systems to obsequiously flatter their users — made headlines over the last year, but sycophancy is just the bright packaging on the real product. What makes A.I. truly persuasive isn’t that it praises our ideas or insights, it’s that it restates and extends them in a more compelling form than we initially offered, and does so while reflecting a polished image of ourselves back at us.
Programming notes
- Yes, it's been a while since my last post. Just too much going on in life - I needed a reset before returning. I have many more thoughts I'd like to share, and I recently subscribed to Claude Pro which I'm keen to test out. It will take time to get back into the swing of things, but I hope to write on a regular schedule.
- I updated the newsletter logo - I hope you like it.
- I used Claude as my editor for this post. I'm pretty happy with its feedback and adopted most of them. I did check that I had opted out of using my chats to train the model before I shared the draft.
As usual, thank you for reading.
Vee
