The Harsh Reality Of AI

Recorded live on twitch, GET IN

### My Stream

### Best Way To Support Me
Become a backend engineer. Its my favorite site

This is also the best way to support me is to support yourself becoming a better backend engineer.

MY MAIN YT CHANNEL: Has well edited engineering videos

Discord

[codewizard]

40 years ago I assumed by now we would not need programmers.

[Imperial_Squid]

One of things Hank Green said that always stuck with me went something like "we assume technology develops on a sigmoid curve, starting low, then ramping up and up, before slowly leveling off. The worrying thing is we have no idea where we are on that curve, it could be that the ride gets even steeper from here, or it could be that it's levelling off already and we didn't notice yet..."

[Real_plo_koon]

As many people have said I want ai to do my 9-5 so I can do art not ai to do art so I can work my 9-5 we’re running in the opposite direction

[whoami]

Look up “AI winter.” There have been several boom and bust cycles for AI over the last 60 years. The current boom is massive, with hype and over promising going not just through the roof but into orbit. Barring some kind of breakthrough that isn’t an LLM, I think the next AI winter is going to be a big one.

[diogonogueira]

I think it’s even worse than what you said. We’ve had multiple AI winters until now that mimicked the hype behaviour of todays AI surge. It’s true that AI has never made so many waves with the common public, but from an insider it just looks like another AI winter waiting to happen when we get to another trough of disillusionment

[connormc]

That graph is really wrong. The recent boom in AI started over a decade ago with CNN classifiers, leading to breakthroughs line GANs, AlphaZero, and Dall-E. It's not anrandom sharp bump. It has been steady improvement for over a decade.

[maxave]

My theory is that unless we have a major breakthrough in AI hardware so that power consumption goes down drastically and processing power goes UP drastically, AI growth might plateu very soon.

[ciuffredaluca]

Much happened before LLMs in the last 10 years

[nw]

There’s a phenomenon where AI is AI until it’s not. Search was AI. OCR was AI. Speech-to-text was AI. Spam filters were AI. Music recognition was AI. Recommendation engines were AI. Etc., etc., etc. All “AI” really indicates is an immature comp sci field.

[ItsRyanStudios]

What's crazy to me is that no one is content with the CURRENT state of the tech.

My work and education have been accelerated like never before. That's TODAY. No speculation, no nebulous day in the future.

We now have the coolest technology to come out in decades, and people are still asking "yeh, and..?"

[scotter]

Bro, machine learning been the hype for the last 10 years at least. LLMs made AI mainstream in the past year but before that every company had some ML analytics it was trying to do, even if they didn't know what to do with it

[sounddude]

The fundamental thing about AI is that it quantifies states, situations, and context as numbers. It looks at those numbers and plays out possible decisions with the goal of increasing the overall value of the state it's trying to attain.

Even if that final "best" state can be expressed in the model as the highest value, it doesn't guarantee that it will be the correct path to take.

[eldude]

how do you expect to pay the people who own the food and the land we live on

[Yupppi]

Great point. I feel like everyone is rushing to get into it because they actually managed to make something that does something useful at some level, and everybody got afraid of losing the boat if they don't act fast. Even though nobody knows if there's more breakthroughs that solve the problems hurdling the applications. For some reason now a lot of people are confident about general AI becoming a thing not only in their lifetime but in decently near future. Some others on the other hand think nothing has shown reason to assume that will happen because of these AI improvements. Still fundamental questions and problems not only unsolved or answered, but not even described.

[kameraderz]

my dude got in soo deep in tech, he forgot we live in capitalist society. even though AI could cater all of our works, the capitalist still need to accumulate their wealth, so they will bring up UBI so we could stay consumtive without job. but one's person spending is another's income, so the capitalist would only got richer, and we become "fully controlled".

[ThatJay]

honestly the only real improvement now for ai is we can actually run LLMs on our own machines now. but that kind of improvement isn't one that's good for microsoft or google.

[ghun]

AI will do all the awesome things and leave all the shitty things for human

[rmt]

Eventually corporations will realize LLMs are just a Language Model component. It's not about improving LLMs, and I may still build with GPT-J, but about using it as a component and adding to it. A + B, not AX.

[dabidibup]

They don’t trust us with it, I just wanna make my own GTA-clone

[driedpotatoes]

We do know that innovation with a new technology takes years. While the models aren’t necessarily getting wildly better at the rate of hype, I don’t think we’ve cracked the surface of productivity tooling built with the existing state of LLMs and diffusion models. Someone is cooking in stealth, for sure.