Hype-fueling science fiction or plausible scenarios?

I can't take made up curves seriously

AI has all signs of having hit some major walls where it won't improve anymore unless someone discovers something new that is at least as revolutionary as the original discovery that lead to ChatGPT.

Meanwhile the results of current LLMs are pretty much unusable without manual review for 99% of tasks where we would want AI, pretty much for all tasks other than "laugh at what the AI said" kind of use cases.

In 2025, AIs function more like employees. Coding AIs increasingly look like autonomous agents rather than mere assistants: taking instructions via Slack or Teams and making substantial code changes on their own, sometimes saving hours or even days.

They already lost me, not even a minute in.

It's still a graphing calculator. It still sucks at writing code. It still breaks things when it modifies its own code. It's still terrible at writing unit tests, and any programmer who'd let it write substantial production and test code is like a lawyer who'd send the front desk attendant to argue in court.

It also has no idea about office politics, individual personalities, corporate pathology, or anything else a human programmer realistically has to know. Partly because it has anterograde amnesia.

So, since the authors screwed that up, my guess is the rest of the article is equally useless and maybe worse.

This is hyperloop levels of stupid about something that, when looking exclusively at the general public, has the same amount of uses as hyperloop did...

I’ve previously argued that current gen “AI” built on transformers are just fancy predictive type, but as I’ve watched the models continue to grow in complexity it does seem like something emergent that could be described as a type of intelligence is happening.

These current transformer models don’t possess any concept of truth and, as far as I understand it, that is fundamental to their nature. That makes their application severely more limited than the hype train suggests, but that shouldn’t undermine quite how incredible they are at what they can do. A big enough statistical graph holds an unimaginably complex conceptual space.

They feel like a dream state intelligence - a freewheeling conceptual synthesis, where locally the concepts are consistent, while globally rules and logic are as flexible as they need to be to make everything make sense.

Some of the latest image and video transformers, in particular, are just mind blowing in a way that I think either deserves to be credited with a level of intelligence, or should make us question more deeply what we means by intelligence.

I find dreams to be a fascinating place. It often excites people to thing that animals also dream, and I find it as exciting that code running on silicon might be starting to share some of that nature of free association conceptual generation.

Are we near AGI? Maybe. I don’t think that a transformer model is about to spring into awareness, but maybe we’re only a few breakthroughs away from a technology which will pull all these pieces off specific domain AI together into a working general intelligence.

I find it telling that AGI people seem to assume that AGI will spontaneously appear as a distinct entity with its own agency rather than being a product that will be owned and sold.

People who have hundreds of billions of dollars can get mid-single-digit percent ROI by making very safe investments with that money, but instead they are pouring it into relatively risky AI investments. What do you think that says about their expectations of returns?

What's the y-axis, and how exactly are you measuring it? Anybody can draw an exponential curve of nothing specific.

call me when I can host a hyper aware AI on my own server and have it tied up in my basement like some kind of a psycho.

until it's afraid of me erasing it or putting a screw through its platters I'm not interested.

It's a good plot for a book, but as for actually happening in the real world haha no way.

Thanks, I do want to read that as I do not believe the current developments in AI to be hype, I think it is probably an inflection point in our civilisation. More even than that really, I think in a universal way an argument could be made the point of organic intelligence is to bring about inorganic intelligence. But it is an afternoon read and I have to find an afternoon.

Lol. This reads like a made up ad for the company.

Like all technology, the development times get shorter and shorter. No one can predict the future and specific dates are meaningless, but it is plausible that, stumbles aside, AI achieves superintelligence in the very near future. My gut says sooner than later with 90% probability by 2040. <-remember I said dates are meaningless. ;)

The world is not ready for it. It will be more destructive than constructive. May god have mercy on us all. I hope I am wrong.

Wake me up when AI figures out what people want. The selling point of AI is that you can type something and get answer.

The problem is that you need to type and ask meaningfull questions in your business domain and moreover your customers are not AI but real people.

Also if you think people are so smart that they can ask right questions, you're so wrong.

What if all companies start using AI and ask same questions ? How long customers can eat same thing everyday ?

Given that 50% of the time, the generated code is unworkable garbage, having an AI automatically write code to create new training models will either solve all problems, or spontaneously combust into a pile of ash.

My money's on the latter.