The Generative AI Con

These companies are heavily subsidized by investors and their cloud service providers (like Microsoft and Google) in an attempt to gain market share. It might actually work - but this situation, where a product is sold under cost to drum up usage and build market share, with the intent to gain a monopoly and raise prices later on - is sort of the definition of a bubble, and is exactly how the mobile app bubble, the dot-com bubble, and previous AI bubbles have played out.

I have a 6 million+ word archive with ChatGPT.

It truly is like having an army of interns, each a confident undergrad in a different subject, who have paid attention to every lecture they ever went to, event the one they'd popped acid just before going in.

It's right more often than it is wrong, but some of its clangers are almost unbelievable.

Yet, having never written a line of code, to build a python application that analyzed election data and applied the results to an interactive map, that gave constituency specific data on hover.

It invariable uses the word clarify instead of correct when challenged. Yet it knows that a clarification is refining an answer within the set of the previously proffered answer, and a correction is a revision on an answer outside of the set previously provided.

It believes that this is so consistent that, on the balance of probabilities this is coded and not purely as a result of training data.

When asked to write an article on this, and include the instances from that conversation where it had incorrectly used the word clarify, it edited the quotes to remove the evidence (probably the most egregious act I've witnessed it perform).

I still use ChatGPT, even more so now since DeepSeek got slow, but I watch it like a hawk.

I still call it out every time it prevaricates or flat out lies, it still promises to do better, it still, on being challenged, acknowledges that these assurances are dangerous lies to anyone who doesn't know it's lying.

But, for me, it is still a highly useful tool.

It frequently makes assumptions that would be made by those in a field I am unfamiliar with in a way that allows me to refine arguments.

Sharing ChatGPT chats can be a very helpful means of sharing one's thought process.

I have it on strict instructions not to create unless specifically told to, to not regurgitate what it already has, to focus on critiquing instead of echoing or praising.

Yet it still reckons 70% of its output violates these instructions.

But the remaining 30% justifies the time I spend using this remarkable, next generation, automation machine.

Because that is what it is.

*To say it is intelligent is like saying a scanner has an eye for detail. Yes, a scanner identifies every pixel but and LLM is no more a brain that a scanner is an eye. (And, yes, I know, but this is a line for people who don't know the neurological processing behind sight, which to be fair, is frequently not very logical.)

So it is a threat to people who earn money on fiverr writing bits of code or designing logos - hell yes.

It is a threat to those who code complex systems or who's designs can add actual digits to market share? hell no. Or at least not for the foreseeable future.

Just as the dotcom bubble funded the internet infrastructure that we still use today (just very inefficiently), it is unlikely these trillions will be completely wasted

Even the US Government is getting involved in subsidizing these companies and all of the infrastructure and resources needed to keep it expanding. We can look forward to even more methane power plants, more drilling, more fracking, more noisy data-centres sucking up fresh water from local reserves and increased damage to the environment that will come out of the pocket books of... ?

Update: And for what? "Deep Research"? Apparently it's not that great or world-changing for the costs involved. It seems that the author is tired of the yearly promise that everything is just a year or two away as long as we keep shovelling more money and resources into the furnace.