GPT-5 Launch Day is NEAR | The End of Privacy and 'Digital People'

The latest AI News. Learn about LLMs, Gen AI and get ready for the rollout of AGI. Wes Roth covers the latest happenings in the world of OpenAI, Google, Anthropic, NVIDIA and Open Source AI.

My Links 🔗

#ai #openai #llm

#ai #openai #llm

BUSINESS, MEDIA & SPONSORSHIPS:
Wes Roth Business @ Gmail . com

Just shoot me an email to the above address.

[andreaskrbyravn]

It kind of piss me off that they are light years ahead of models and we still sit on gpt 4 when they clearly have much better models ready.

[chad]

"Its definitely about to be released. Probably. But its not going to be GPT-5"
Fantastic news as always!

[tomaszzielinski]

If you ever need to find where in the world I am, don't bother using AI - I'm at home.

[Saiyajin]

I believe that OpenAI has found a new way to train their AI and treated it on gpt2. If that’s true, imagine how different will gpt4 or 5 be

[monkeyjshow]

Who suggested that reviews should be censored??? I want the specific names of the would-be tyrants. Censorship is a cardinal sin 6:21

[alwaysyouramanda]

They’re about to dilute us out of existence here online.

[themartdog]

I'm really not excited for more LLMs. They're getting better with each release, but these incremental increases are not getting me out of bed. OpenAI has kind of created this problem for themselves hyping up GPT-5, because if it doesn't do something crazy that no other LLM does, no one will care that much.

[MateoTeos]

Why do we need yet another non-private, highly censored, and biased search? 😅

[phen-themoogle]

One thing I found funny recently is GPT4 and Claude3 both couldn't read a kanji character I sent them via screenshot (was a kanji flashcard app that showed the correct stroke-orders and reading and word examples/sentences with it in tons of context too lol), and although Claude gave a synonym and was close on some words from the sentence, gpt4 was totally off got everything wrong, but then Gemini1 the FREE model got everything 100% correct, like it's ability to process Chinese/Japanese characters is superior to the paid models. OpenAI also launched an office in JAPAN! This kinda made me a little upset, but also made me giggle lol

[danielmartinmonge]

I don't trust anyone who just talks about their products. Right now, we are facing a crucial moment in the industry. Some expert voices are raising concerns about the possibility of AI reaching a ceiling, and the very similar capabilities of every modern model seem to support this hypothesis. GPT-5 is about two orders of magnitude larger than GPT-4 in terms of data and GPUs. The hype driving this insanely expensive journey demands results that blow our minds. Sam Altman is hinting at granular deployments and similar concepts, claiming that he doesn't want to shock everyone. And I wonder: does OpenAI not want to shock everyone, or CAN it not shock us? We'll find out soon. But if GPT-5 is barely superior to GPT-4, that could collapse all multibillion-dollar investments. Who would invest 100 billion in a model that is only slightly better than a model that costs 1000 times less?

[kingthame]

Most people don't actually know that your pictures have metadata attached to them. That's how they are easily tracked.

[DanFrederiksen]

The other day my GPT4 started doing much shorter concise answers and then went back to verbose chatter bot. I've noticed many changes so that to me suggests that they don't really have a significant leap or they would just let GPT4 be and focus on the next. One can also remark that the transformer has been the basis for everything, even Sora seems to use it in the image generation, not just the bridge to natural language. If that premise still holds then the next system wont be a quantum leap. And might not even be a bigger model. Might even be a little smaller. GPT3.5 is smaller than 3 yet significantly better iirc. I hunch that vastly superior intelligence is possible within 1.7T params.

[Crawdaddy_Ro]

We've got to keep in mind that these larger companies (including Meta) are the only ones capable of reaching the next level of AI without significant changes to physical or digital architecture. We're making decent progress, but the development isn't fast enough to ever catch the big fishes. By the time open-source is able to catch up, the big companies will have moved on to the next paradigm. They're thinking several moves ahead because they have to. We'll have what many will call AGI within a year, definite AGI within 2 years, and will be relatively ubiquitous within 3-5 years (just in time for the wave of AI robotics).

[ploppyploppy]

If it's censored or full of other people's 'principles' then we can skip over it and wait for someone to release one that is actually useful.

[DefenderX]

If you solve the safety of AI, you can go as fast as you want.
So I think after GPT-5, I think Anthropic can slingshot past GPT-5.
But also, the capacity needs to be in place as well, which needs capital, so the race is very exciting.

[starblaiz]

My friends and family ten years ago: "Why have you privated all your photos and personal information? You're being paranoid!"
My friends and family today: "Oh... oh I get it now!" 😂

[burninator]

Well, the CIA definitely wants that geolocation id tech now

[misterdudemanguy]

This sort of internal/external visibility will remain, and will become worse. What I mean is the external release will always be less-than, up to the point that if AGI is ever achieved, it wont hit the public release. Not for long after it has been used to tip the scale to the discoverers. I'm not even saying this is wrong per se, but it is going to be far different than how the altruists believe. We are not going to advance into an AGI world together, holding hands and singing. And while gov pretends the first right of inspection is for safety, it's really for capture and kill. Just like the patent system captures and kills tech for national security interests.

But even if it does make it to the public, the hardware to run it will be very expensive and power-hungry, and far beyond the reach of most. Which means that the gatekeepers will always have the option of closing the gate, or crippling whatever they let out of the gate.

[monkeyjshow]

Well-articulated and informative title. That made me want to watch, not that all the superlatives wasn't at least a little entertaining

[jameslincs]

This channel is great! Does make me worry we’re not talking enough about AI safety