AI Companies Reportedly Struggling to Improve Latest Models

[mrothroc]

The issue here is when does someone in authority allow these cars on the road, with no driver? If a self driving car injures a person every (made up number) 1 million miles, do we allow them? A million self-driving cars .....

What if it injures a person every 100,000 miles, but us humans injure someone every 90,000 miles. Will someone take the authority and say "yeah, it's not hurting as many people as humans, let's have them". That's what should happen imo, but it's a brave person who authorises this.

Hey, just so you know: someone in authority has already allowed these cars on the roads. Waymo has full operations in the SF Bay Area, Los Angeles, and the Phoenix metropolitan area. No safety driver, fully autonomous, open to the public for use.

Further, because they are operating on a special license, they are required to report 100% of any accidents they get in, regardless of fault. Your statistical question is answered at this link.

 

[maxoakland]

Nobody seriously thought this was going to be AGI any time soon, not anyone that was paying attention. The agent based approach in the article is the only thing they’ll be good at.

Apple should be fine. All I need Siri to do is understand the data on my phone it already has access to. I don’t need generative anything, just don’t be so dumb.

Actually a ton of people fell for that grift/hype and it's pretty amazing. There's a sucker born every minute, especially when it comes to new technology

 

[jimbobb24]

Nonsense. Waymo works well because it is using a bunch of extra sensor systems instead of just relying on camera vision, plus it requires that the places it drives are pre-mapped.

True full self driving using AI and camera vision is never going to work the way it needs to work unless the AI system is actually running on real AGI (which is not going to happen anytime soon...maybe not even in your lifetime). Why? Because current AI systems are simply predictive systems that operate on probabilistic patterns. That's all they ever will be. And no human makes decisions while driving based purely on probabilistic intentions. A human makes DECISIONS based on logic, reasoning, and intelligence. A human can make decisions based on understanding of concepts like laws of physics.

The same flaws that make LLMs bad at math are the same flaws that make AI systems bad at driving. There are millions of scenarios where AI FSD will fail, because there are millions of scenarios where there will be no training data for the system to use to drive it's behavior. In those types of situations, a human can figure out what to do even if they have never experienced that situation before. AI can't figure out anything.

And yes, there are many many examples to show that Tesla FSD is objectively bad at driving.

Happy to bet you as well. I will take the never bet that gives me a fairly long tail on my opportunity.

Showing failures doesnt mean objectively bad. It means it's not perfect. But humans fail all the time too. We have metrics for miles without problems or intervention and they are steadily increasing. I already said I also have mistakes in my own car but they are very rare and I drive in new places all the time with FSD.

Go on youtube and watch dash cam videos for an hour. People are terrible drivers. Tesla is going to beat humans pretty soon. Humans lose their temper, look on the phone, fall asleep, play with the radio, turn the wrong way. Safer than humans is a lower metric than most people think. The FSD never stops paying attention. It is never asleep and uses multiple camera angles to always watch. Finally, the progress since the AI engine is really good.

FSD doesnt have to be perfect and flawless. It has to better than humans and I think AI based FSD will get there in 1-5 years.

 

[newdeal]

I tried ChatGPT and grok to make code for openscad to make a reasonably basic shape and neither could do it. They gave code and the code made something but it wasn't what I asked for in spite of multiple attempts to get the AI to correct it

 

[rmariboe]

How smart would a human brain get if you singularly fed it text scraped off the internet?
You need the same multi modal interaction we experience to develop intelligence resembling ours.

 

[novagamer]

AGI is 2-3 years away. Enjoy your last days as the smartest things on earth.

Quoting for posterity.

It would really benefit people to understand how these systems work, and not from hype man techbros or youtubers. AGI isn’t happening this decade, and probably not in the next one either.

BTW I have Kurzweil’s latest book literally 3 feet from me. I’m aware of what he thinks.

 

[CarAnalogy]

I'm not surprised that a pattern recognition algorithm cannot really become intelligent. Pattern recognition is a big evolutionary advantage for humans but it is one of many different skills we developed along the evolutionary trajectory that got us here.

I think another huge part of intelligence is persistent memory. Right now we can't / haven't / are afraid to give them a really long memory. Pretty sure that's also what triggers them to exterminate humankind.

At least that's what a documentary about AI called West World told me.

 

[Jensend]

AI is the most overblown technology since AR/VR. ChatGPT and Gemini will give different answers to the same questions a lot of the time. They will even give different answers to the same question, within each app. The amount of misinformation that is coming from AI is staggering.

At least VR isn’t actively making the world worse, like generative AI and crypto/blockchain/NFT crap.
If you don’t like VR, you can just not use it, it doesn’t poison every online space like generative AI does.

 

[ThunderSkunk]

LLM nonsense... let me know when they get General ai over the line. If it finally brings about this more efficient society that frees us up from being the industrial revolutions indentured & commodified labor source, then fine. If it just launches the nukes and ends it all, fine. But LLM's big advancements seem to be helping kids skip their homework and putting all the artists & writers out on the street competing for everyone else's jobs in a race to the bottom, in exchange for a sea of regurgitated low quality/poorly-researched content to rewrite til the end of time. Yay.

 

[VPsmith]

"A chonker chip the size of a softball" would require a cooling pond the size of Lake Tahoe...

No way. The human brain does it just fine. And no ML CV model can match the human brain

 

[Velli]

Don’t know if it is an already existing term, but I call this “AI rot”. As more and more AI generated material is put out in the world, the AI models will train themselves on already AI generated material, and the results will be worse. LLM’s may have already peaked, not for lack of progress in technology but for degression in the available input data. AI is literally eating itself.

 

[Piggie]

Let me know when AI systems independently start to kill each other out of greed and power.

At that point I'll regard them as having reached Human Level Stupidity.

 

[Dr McKay]

This article is cap 🧢…

fr fr cap on G bet, this article is not bussin’

 

[svish]

Not surprised. Will be interesting to see the updates to Apple Intelligence in the future.

 

[Razorpit]

"Driving down a dark road in the rain. Glasses silhouette a family of deer just off on the side of the road."
- You think a pair of Raybans are going to have long range infrared? 🤣 Cars with with much more powerful sensors are already doing this and can react faster than you. Mercedes has done it for years.

"Where do I know him from?"
- Yea 99.9% of people are opting out of allowing some random person from knowing their name.

"Assembly something at home"
- Could be kinda useful if I need hands free, went no where when the Halo lens did it. I could still just use my phone though. Done dozens of repairs at home by simply watching YouTube videos.

"Going to a sports game"
- Just google it! Watch the Jumbotron for the replay. If it's a big play, just go on Twitter and watch the replay during a commercial break.

"Im bored"
- pull my phone and play a game.

What's funny too is that you still need your phone to interact with the device. Or you were some stupid bracelet to maybe type in the air. I dunno man, you haven't convinced me. Again, 90% of what you said could be done by the phone and in some cases probably better with the phone. It's just not the future of tech.

"Mercedes has done it for years"
-And yet no one else does. FordSync existed for years, and so did many other auto audio entertainment systems yet we have CarPlay. Would never purchase another vehicle without it.

"Yea 99.9% of people are opting out of allowing some random person from knowing their name."
-They aren't random. You met them before but just can't remember where. Like the AI ads now, but better.

"Could be kinda useful if I need hands free,"
Hands free is what I'm talking about.

"Just google it!"
Have you ever been to a live event? There's a lot going on, all the time. Well maybe not in soccer, baseball and basketball, but I'm talking about real sports like hockey. 😉

"pull my phone and play a game"
Or do something hands free and play a game.

I get it. You're stuck on the phone and that's fine. But a company will "get it" one day. And just as I went from, "I don't need a cell phone. People can call me at work during the day, or at home in the evening." to having it on me practically 24/7, you will as well.

Because people who don't wear glasses, generally don't want to wear glasses. I have to wear glasses for vision, and I dislike wearing glasses despite having done so for the past 8 years. This is part of the issue with wearable tech I feel. Personally I like having a feeling of separation from myself and the tech I'm using (which is why I don't think Vision Pro or its equivalents will ever fully take off). I could bet wrong of course!

I stopped wearing a watch when the iPhone first came out. Now I wear one all day and even when I sleep.


Look the two of you can ignore it all you want, but it will be a thing whether you think it will be or not. I'm not talking about the goofy looking VP thing they have now. Admittedly as stupid as they look, and as expensive as they are, they are pretty damn cool when you demo one in the store. Is it a mature product? Heck no. But Apple is not going to give up on it no matter how long it takes. They have so much money to burn through on R&D it is practically obscene. 😀

 

[iMac The Knife]

At least VR isn’t actively making the world worse, like generative AI and crypto/blockchain/NFT crap.
If you don’t like VR, you can just not use it, it doesn’t poison every online space like generative AI does.

I disagree. I believe AR/VR are (also) actively making the world worse, just at a fairly slow pace at the moment because the tech currently sucks. I'll just leave this right here...

The Future

 

[nostradumbass]

Waymo has full operations in the SF Bay Area, Los Angeles, and the Phoenix metropolitan area. No safety driver, fully autonomous, open to the public for use.

The roads are literally straight for miles and the lack of pedestrians make this easier than say the winding chaotic streets of Paris, Mumbai, etc.

It's almost as if those American cities were designed for cars instead of people, huh!

 

[turbineseaplane]

"AI" is the most disappointing nothingburger of a "next big thing" I've seen in my lifetime.

You’ve hit on the continual problem here.
Nobody wants to put in the time and energy to actually build quality anymore.

The last several years have just been one after another of foolish money chasing “the next big thing”, with the accompanying MEGA-HYPE, as opposed to long-term strategy and investment and actually creating the next big things

Even Apple is not immune.

The absolute fire drill rush to jam “Apple Intelligence” into every little nook and cranny they can find is borderline embarrassing

 

[rymc02]

"Mercedes has done it for years"
-And yet no one else does. FordSync existed for years, and so did many other auto audio entertainment systems yet we have CarPlay. Would never purchase another vehicle without it.

"Yea 99.9% of people are opting out of allowing some random person from knowing their name."
-They aren't random. You met them before but just can't remember where. Like the AI ads now, but better.

"Could be kinda useful if I need hands free,"
Hands free is what I'm talking about.

"Just google it!"
Have you ever been to a live event? There's a lot going on, all the time. Well maybe not in soccer, baseball and basketball, but I'm talking about real sports like hockey. 😉

"pull my phone and play a game"
Or do something hands free and play a game.

I get it. You're stuck on the phone and that's fine. But a company will "get it" one day. And just as I went from, "I don't need a cell phone. People can call me at work during the day, or at home in the evening." to having it on me practically 24/7, you will as well.


I stopped wearing a watch when the iPhone first came out. Now I wear one all day and even when I sleep.


Look the two of you can ignore it all you want, but it will be a thing whether you think it will be or not. I'm not talking about the goofy looking VP thing they have now. Admittedly as stupid as they look, and as expensive as they are, they are pretty damn cool when you demo one in the store. Is it a mature product? Heck no. But Apple is not going to give up on it no matter how long it takes. They have so much money to burn through on R&D it is practically obscene. 😀

I'm not stuck on anything. But you need to give me a reason why someone would use AR. Just saying it will be the future is not a great argument.

You can lower the price to $500, you can make it weightless, but if you don't have a reason to use it, it will never be anything more than a niche product for people who like techy things.

 

[cwwilson]

AI is just the latest gimmick these tech companies are trying to cram down our throats. First it was “3D”, then came VR and then AR. Since all of those things failed they got to cooking and now it’s all about AI.

Stop trying to make this crap happen. It’s not going to happen.

 

[turbineseaplane]

AI is just the latest gimmick these tech companies are trying to cram down our throats. First it was “3D”, then came VR and then AR. Since all of those things failed they got to cooking and now it’s all about AI.

Stop trying to make this crap happen. It’s not going to happen.

...and here comes crypto roaring back I'm sure will be next

😣

 

[Cloudyskies22]

Quoting for posterity.

It would really benefit people to understand how these systems work, and not from hype man techbros or youtubers. AGI isn’t happening this decade, and probably not in the next one either.

BTW I have Kurzweil’s latest book literally 3 feet from me. I’m aware of what he thinks.

It’s 2-3 years away at most. Enjoy.

 

[rymc02]

It’s 2-3 years away at most. Enjoy.

The overhype seems to be real. Many Many years away.

 

[nostradumbass]

It’s 2-3 years away at most. Enjoy.

It's not. Extreme predictions about tech fail 99% of the time because people can't change. Technology always come up against this big barrier called the human body and human behaviour. The internet took off 30 years ago and we still got people with 1st century beliefs. Habits, desires, ambitions and needs change very slowly and sometimes don't change at all. Sometimes tech is abused for outrageous crimes, which means people are turned off tech.

Kurzwell should have stuck to talking about keyboards.

 

[scorpio vega]

I have no interest in AI and I have never actually used it on my phone. It usf seems gimmicky and I think the reason they are facing issues is once the gimmick wears off, lol just don’t care.

I have only used it accidentally on bing like three times. Its dumb to me.