this post was submitted on 14 Sep 2024
23 points (79.5% liked)

Videos

14271 readers
287 users here now

For sharing interesting videos from around the Web!

Rules

  1. Videos only
  2. Follow the global Mastodon.World rules and the Lemmy.World TOS while posting and commenting.
  3. Don't be a jerk
  4. No advertising
  5. No political videos, post those to !politicalvideos@lemmy.world instead.
  6. Avoid clickbait titles. (Tip: Use dearrow)
  7. Link directly to the video source and not for example an embedded video in an article or tracked sharing link.
  8. Duplicate posts may be removed

Note: bans may apply to both !videos@lemmy.world and !politicalvideos@lemmy.world

founded 1 year ago
MODERATORS
Flutter@lemmy.world
Ghostalmedia@lemmy.world
can@sh.itjust.works
qaz@lemmy.world
qaz@lemmy.blahaj.zone
23
OpenAI’s new “deep-thinking” o1 model crushes coding benchmarks. (odysee.com)
submitted 1 month ago by 101@feddit.org to c/videos@lemmy.world
27 comments fedilink hide all child comments
top 27 comments
sorted by: hot top controversial new old
[–] simple@lemm.ee 19 points 1 month ago* (last edited 1 month ago) (1 children)

Outside of benchmarks it's really not as big of a deal as openAI wants you to think it is. In most cases it's slightly better than Claude, except it uses 50x the tokens repeating info to itself and is way slower. There are a lot of people that tried o1 online and are posting screenshots of it making basic mistakes or gaslighting itself in its chain of thought. Not to mention you only get 30 messages PER WEEK since it's such a waste of energy.

It's a desperate attempt by openAI to stay relevant now that competitors and even free models are catching up.

[–] scrubbles@poptalk.scrubbles.tech 3 points 1 month ago

I've learned all AI marketing is greatly embellished, focusing only on best cases. Sure, it can solve a coding problem, if you work on prompt for an hour, give excruciating detail, and run it a few dozen times

[–] just_another_person@lemmy.world 5 points 1 month ago (1 children)

Eh. If it what they say it is, it's still not going to replace a solid engineer. This is also still not capable of actually developing novel reasoning and logic solutions, so the threat of not being able to own the IP is still a thing. There's also still this bullshit of OpenAI hinting that they want to retain rights to created works through certain licenses somehow?

Fuck these assholes up and down the aisle.

[–] Lauchs@lemmy.world 5 points 1 month ago (1 children)

Why?

A half decade ago we would've laughed at a machine passing the Turing test...

[–] just_another_person@lemmy.world 1 points 1 month ago (1 children)

Why what?

[–] Lauchs@lemmy.world 5 points 1 month ago (5 children)

Why could it not replace an engineer?

The previous limits of technology exploded less than half a decade ago, seems wild to assume that's the end of that kind of growth.

[–] vinnymac@lemmy.world 5 points 1 month ago* (last edited 1 month ago)

Eventually, we might get there, sure. But I don’t see any reason to believe this is it, and I use AI to assist in my programming every day.

If you instead said, some engineers will be replaced by AI. I’d definitely agree, and without a doubt they’ll try, repeatedly.

[–] lemmydividebyzero@reddthat.com 4 points 1 month ago* (last edited 1 month ago)

In its current state?

Writing code is a small part of being a software engineer and compared to those coding tasks with very detailed instructions about the input, constraints and the output (even with examples), actual tasks are usually missing lots of information you need to find out from different people and there is a huge code base that can't be transfered to the model.

If it can fully replace a software developer, it can replace almost anyone's job.

[–] Floey@lemm.ee 2 points 1 month ago

Technology is always progressing but nobody can say what the next big thing will be, if you really think you are that prescient you can make loads of cash predicting things. Companies are hungry for the next big thing though and will do everything to convince us that they have it, AI is an enticing grift because it's so misunderstood. The next big thing wasn't AR or VR or the metaverse, and I don't think it's going to be generative AI either, it's already plateauing and not profitable, even with billions of dollars behind it.

[–] Valmond@lemmy.world -1 points 1 month ago (1 children)

Sure, make it drive a car first, a thing 99% of the population can do before attempt coding ^^

Coding is actually quile complicated, especially in old existing codebases. Add that they train them on any crap code they can find...

[–] stephen01king@lemmy.zip 1 points 1 month ago (1 children)

No way are you going to convince me 99% of the population can drive. Go get a more accurate statistics before trying to use it to dismiss something.

[–] Valmond@lemmy.world -2 points 1 month ago* (last edited 1 month ago) (1 children)

It's an example, and most adults (if I have to explicit it) can drive, or can learn how to.

Coding not so much.

So AI that can't even drive, can code suddenly? I don't think so.

Better like that?

[–] stephen01king@lemmy.zip 2 points 1 month ago* (last edited 1 month ago) (1 children)

Most adults can also learn to code, if they actually tried. If you're gonna add the argument that most people can't code proficiently, most people can't drive proficiently, either.

Also, driving and coding are completely different set of skills that it's kinda worthless to compare them. Some people can code just fine but might never learn how to drive because they didn't need to, so to consider driving as a prerequisite skill to coding doesn't make sense.

[–] Valmond@lemmy.world 0 points 1 month ago (1 children)

Well I think you're wrong here, and about any adult can learn how to drive, but only a small subset can learn how to code. Not learning how to throw a simple script together, real codeing.

Coding is engineer level, engineers build cars, they dont only drive them. For me the difference is the same between a developer of a software and the user of said soft.

One it way way way more complicated, and IA is supposed to do that "soon" when it can't even drive a car.

Nah, not happening any time soon.

[–] stephen01king@lemmy.zip 2 points 1 month ago (1 children)

I think you're completely wrong by still comparing skills that have no relation to each other. What's the similarity between driving and coding that would require an LLM to be need to do one before you can believe it can do the other? Explain that leap in logic properly before you continue with your argument.

An LLM is designed to output text. Expecting them to drive to prove their ability to output code is like expecting them to dance to prove their ability to produce poems. It's inability to do an unrelated skill has no bearing on it's ability to do a different one. You're basically judging a fish on its ability to walk on land, and using that as the basis to judge its ability to swim.

[–] Valmond@lemmy.world 1 points 1 month ago (1 children)

Neural networks are quite similar in complexity, whatever they output.

Driving is way less complex than programming.

[–] stephen01king@lemmy.zip 2 points 1 month ago (1 children)

What does that even mean? Neural networks have varying levels of complexity, even within the same technology. Even the same LLM model can have different number of tokens that differentiate the complexity of their operation.

So instead of using a neural network that is designed to input and output text and making it learn to output coding, which is also text, you think it's supposed to be easier for them to make it instead analyse various video and audio input from multiple cameras, and then output the various actions that is required for it to drive a car? Does that make sense to you?

[–] Valmond@lemmy.world 0 points 1 month ago (1 children)

They're all based on the same tech, text, images whatever is just raw bytes. Train it for turning left, detect cats or whatever, same battle. It's just harder when the problem is complicated like driving, and more so when programming.

[–] stephen01king@lemmy.zip 2 points 1 month ago (1 children)

Doesn't sound like you understand it very much, like I thought so.

[–] Valmond@lemmy.world 0 points 1 month ago* (last edited 1 month ago) (1 children)

Lol Dunning Kruger at its finest, with a sprinkle of insult lol.

Call me back when your PC can drive, or, lol, replace a developer.

[–] stephen01king@lemmy.zip 1 points 1 month ago (1 children)

Oh, you actually know what your problem is, good job.

[–] Valmond@lemmy.world 1 points 1 month ago (1 children)

Insults will not let you win, what a loser approach lol

[–] stephen01king@lemmy.zip 1 points 1 month ago (1 children)

Says the guys who started throwing the insult. It's nice that you are kind of self aware of your own shortcomings.

[–] Valmond@lemmy.world 1 points 1 month ago (1 children)

Call me back when your PC can drive, or, lol, replace a developer.

[–] stephen01king@lemmy.zip 1 points 1 month ago (1 children)

And keep repeating those words until you're satisfied. Maybe you can finally understand why they're meaningless.

[–] werefreeatlast@lemmy.world -1 points 1 month ago

....officer! I told the car to not hurt anyone! I can't help it if he's sarcastic!