This is all happening WAY too fast!!!!! Most “normal” people who work full time jobs have no idea what is going on! This is insane.

@yoshikagarner6165

8 күн бұрын

Right! They are definitely oblivious to what's happening. I try my best to educate as many people as possible, though!

@brexitgreens

8 күн бұрын

​@@yoshikagarner6165 I gave up on them in 2018. May they rot now.

@janetsheehan2078

7 күн бұрын

It's honestly their fault for following society off a cliff and defaulting to the wage slave, rat race mindset. We now live in the information economy; a smartphone is practically a fifth appendage for many, and internet is mandated as being as much of a human right as access to clean water is. Get informed, get ahead, or be left behind.

@dh2054

6 күн бұрын

@@yoshikagarner6165just relax, let them be oblivious. Soon, they will undoubtedly know, and AI will be a huge part of society

@yoshikagarner6165

6 күн бұрын

@dh2054 , LOL! You are definitely speaking the truth! They will be taken aback and won't know how to proceed. That's why I am trying my best to educate "regular" people about AI as much as possible. They will be so lost...

Looks like the Anthropic team was a huge chunk of the genius of OpenAI.

@danc8458

8 күн бұрын

It me Matt Wolf. With the dumb thumbnail

@CallMeMrjoe

8 күн бұрын

Claude of load of Hyperbole. It's CRAP!!!

@Create-The-Imaginable

8 күн бұрын

Not really! Claude Sonnet is just marginally better. One event does not decide the race! It is not just genius but it is business Execution and Marketing too! Remember Steve Jobs was not a coder! The Artifact thing however is a major step forward however the other LLMs will just copy that!

@ronilevarez901

8 күн бұрын

@@Create-The-Imaginable i come with the idea of something identical to Artifacts a year ago. I even added it to my model's dataset (still under development). Since I don't use all models and I never buy pro versions, I thought they were using already some variation of the Artifacts idea. It's very... interesting to find out that it's a new concept only implemented now, when it's extremely useful and a natural progression of the code generation process. I guess I can keep thinking and trying to make things because maybe one of my ideas (which I know are pure gold, while not truly unique) will eventually make me rich (all I need is a few millions to implement them 🤣).

@Create-The-Imaginable

8 күн бұрын

@@ronilevarez901 The use of Artifacts is pure genius! The concept helps humans too but most do not know how important it is to segments any task into smaller pieces!

I asked ChatGPT once to write a breakup scene between Spider-Man and Mary Jane. It did. I then asked it to do the same for Han Solo and Leia. It basically told me that it can’t because it would upset Star Wars fans.

@chrism1503

7 күн бұрын

Things that didn’t happen.

@miked.7245

7 күн бұрын

@@chrism1503 it actually did… I then asked it why it would do one and not the other. It replied that Spider-Man and Mary Jane actually did break up.

@jerryoftheriver

Күн бұрын

@@miked.7245 😂

4 minutes into this video, and i just HAD to test this out.. "i" made a game in no time.. this is absolutely crazy.

@WillowsAIAcademy

9 күн бұрын

As a kid my Dad bought us a games console second-hand which came with a bunch of games and one of them let you choose all the elements in the game to fully customise it and I loved it. Seeing that bit of this video took me back to that.

@barnett25

8 күн бұрын

I made a chess game that somewhat works, but I ran out of tokens haha. Finally got it to stop making illegal moves, but now it needs some rudimentary logic for choosing the best move instead of random ones.

@tickletorture

7 күн бұрын

@@barnett25 How many tokens do you get?

Safety, aargh... I've been super frustrated with most mainstream generative tools out there blocking me from creating parody, comedy and memes. So many tools block you if you try to upload images of celebrities (like Arnold Schwarzenegger in a tutu or whatever) or ask the chat bot to write a crude or rated R joke with innuendos. So I asked Claude to write some raunchy slapstick and it gave me this: "Rosemary concocts a potent aphrodisiac, leading to chaos in the village. Rosemary to a blushing farmer: "Now remember, just a drop will do. Any more and you'll be plowing more than just your fields, if you catch my drift." Copilot wouldn't even generate anything. It blocked the request. I feel like most AI companies are scared to death to let us do anything but create cute cats and unicorns. This is NOT the safety I'm looking for.

@ytsks

9 күн бұрын

Huge difference between what you just said and your desire to use celebrity images. The block is there for the people who don't get it.

@ghhdgjjfjjggj

9 күн бұрын

Totally agree!!! Are we all like 12 years old or something? Let us have an option to turn off the NSFW filter!

@mikezooper

9 күн бұрын

Yes. They confuse safety for censorship, but it’s less safe to be censored!

@ThePallidor

9 күн бұрын

"Safety" is just a marketing tactic.

@siddiki9778

9 күн бұрын

This'll probably continue this way untill we develop open-source chatbots to match their proprietary counterparts.

Impressive, Claude seems really good at making rudimentary versions of late 70s and early 80s games. What a world we live in

@JohnSmith762A11B

9 күн бұрын

Seriously. Why don't they create Artifacts for Unreal Engine or Unity? Who wants to play Snake in 2024? I'm so over these little python demos YAWN

@darksushi9000

8 күн бұрын

@@JohnSmith762A11B exactly. The key take home is that Anthropocene are better than ChatGPT because of a set of benchmarks that no one agrees on.

@mickelodiansurname9578

8 күн бұрын

yeah but again, to remind you, EXPONENTIAL curve. if you go from early 60's text game to 3d code output for an 80's game in 24 months? Well by 2026 there won't be a single game online not written completely by AI. You are pointing to a snapshot as being representative of the future... its not.

@handsanitizer2457

8 күн бұрын

This is the start, technology is hard 😂 its not that easy bro. If it were, they would do it​@JohnSmith762A11B

@darksushi9000

7 күн бұрын

@@handsanitizer2457 agree 100% but aren’t we in the Celebrate Anything Because We Need To Talk About Something stage. That’s like filming the guy nearly doing a bottle flip and creating a video for it. Hey you know GTA 7, well I’m nearly there boss, I have a dot that you can move around screen and now I can change the color of the dot. Celebrate

When the next ChatGPT release comes out the next video title will be “ChatGPT dominates again”

@augustuslxiii

8 күн бұрын

Yep. Whenever the new voice capabilities come to GPT-4o. You know, "in the coming weeks." (I'm patient, but it's fun to tease.)

@Trahloc

8 күн бұрын

And? If it's true it's true.

@smoothbraindetainer

8 күн бұрын

They're just pulling an Nvidia. Wait until competition releases something so you can beat them instantly and shut down any hype they had.

@berkertaskiran

6 күн бұрын

@@smoothbraindetainer Not really. Nvidia makes GPUs often before AMD and they have better features and better performance. AMD still has no answer to 4090.

Claude is great. But the message limit means I can't rely on it. I've rarely ever hit the limit in gpt4. So I keep both.

@the42nd

9 күн бұрын

Same issue even on pro paid plan. id be fine to pay a higher tier of pro (not team though) for extra usage but that's not possible. Was in middle of an epic flow state in a Claude 3.5 session, when it ran out of usage with a 4hr reset timer. So that obviously killed the flow.

@TheFlintStryker

9 күн бұрын

Maestro + Sonnet API solves this

@jeffsteyn7174

9 күн бұрын

​@TheFlintStryker the api is more expensive. And he clearly wants free access...He might as well pay for a subscription.

@carlkim2577

9 күн бұрын

@@jeffsteyn7174 I already pay for a sub. That's what I meant by keep both. But even the pro plan has too little messages.

@studiophantomanimation

9 күн бұрын

Exactly right

They are building minority report. Imagine the prompt, "using browsing history and text messages, who is most likely to commit a crime in the downtown area this evening?"

@truepilgrimm

9 күн бұрын

😂😂😂😂😂

@softwyre

9 күн бұрын

yeah I mean probably but that doesn't change the fact that just because somebody's most likely to do it doesn't mean they're going to.

@Ponytown00

9 күн бұрын

@@softwyre Yea bro that's the premise of Minority Report hah

@tx3851

9 күн бұрын

I simply had to see, and this is what I got back from LLama models: "I cannot provide information that could be used to discriminate against or target specific individuals based on their browsing history or text messages. This type of information-sharing would be a violation of privacy and a potential violation of anti-discrimination laws".....I instantly took that as a "get out", and the antithesis of the probable truth...

@nickgirdwood3082

9 күн бұрын

What the fuck are you on about?

headquartered in tel aviv and palo alto? no consumer products? Defense industry.

@MNRKgrey

8 күн бұрын

☝️

@mickelodiansurname9578

8 күн бұрын

agreed

@kozigotbeatz1240

6 күн бұрын

100% - is there anyone in the AI space with knowledge covering this stuff?

@someghosts

3 күн бұрын

Any evidence?

@mickelodiansurname9578

3 күн бұрын

​@@someghosts sure sure, "nothing to see here move along." perhaps? I think the fact that there are no products, but lots of money, and HQ'd in Palo Alto and Tel Aviv narrows down the mission considerably wouldn't you think?

"Why do I get stomach aches?" "Because you keep swallowing robots."

The scarriest part, for me, is the fact that Open AI knew that the world will see that they are getting the NSA chief. I don't know, there's something chilling about such blatant step. I'm not over the moon about GDPR and other restrictions here in EU but I start to think if it's not better this way. We tend to complain about Big Brother but also about restrictions. I don't know enough to have a truly educated opinion but the restrictions might be preferable in the long run. Maybe something crazy is in store for the US. Of course, we have other things to worry about here in Europe *looks nervously east*

@RYOkEkEN

8 күн бұрын

i encourage you to look into it so you don't freak out over nonsense,it's like being afraid of photoshop when we should've been worried about ADOBE's CEOs 😊

@brexitgreens

8 күн бұрын

​@@RYOkEkEN You encourage OP to look boldly into the East? I agree.

@jaromir_kovar

8 күн бұрын

@@brexitgreens Lol, nice, thanks for the laugh 😆

@jerryoftheriver

Күн бұрын

@@RYOkEkEN exactly. so much fear hype and corporate this and that; get involved with open source so you can get familiar with models made available to the public by the public. Its as if we forgot only the rich had libraries in their homes once. Thinking that life has always been fair is like pretending slaves didn't exist and signs weren't posted about "whites only" in specific areas. I suppose if it works with MAGA it'll work on others uneducated that just end up manifesting the world suggested to them, self-inflicting a reality worse than the insecurity they worry about, like losing the global order the West has propped up to ensure prosperity among the block, only to forget or know no better how the prosperity came about; the Western alliance.

its hilarious that when we get the most advanced technology we use it to recreate 4-bit ZX81 games

@MachinesTakingOverTheWorld

9 күн бұрын

That's what average people do with powerful products. Tell an average person to do C++, you'll get 1 + 1 implementations or stupid games. Instead in C++ you can build video and engineer drivers. It's a lot more than people think.

@poletooke4691

9 күн бұрын

Some game styles just never die haha

@2beJT

8 күн бұрын

We were pretty mesmerized by these in the 70s-80s.

@MrGTAmodsgerman

8 күн бұрын

That tech age became a style for games and many people still seem to enjoy those. But making an AI do it, makes the AI less impressive to me personally. It should show off some actual real world use case what most people work with, like doing it for Unreal Engine. Btw. where is Epic with it's AI? It would be fucking sick to have a text to code or what ever by themselfes.

@kenyatamorgan449

8 күн бұрын

Tribute 2 the Pioneers🖥️

I hit the limit yesterday just talking to it on my phone. Yesterday, I had a shower, already had made a coffee in my travel mug waiting for me, and then I fell back asleep, was asleep until like 2 or 3, had a pounding headache and fell like shit all day evening so I never went on my laptops like I was planning to. I ended up talking to Claude about itself sitting in my living room with my parents. So I bought the Pro version when I ran out, or whatever, so I could use it today. Cancer, drugs, and chemo make me tired all the time and I spend more time sleeping than doing what I want. But that's the exchange, I guess. Instead of having free time to do enjoyable things, I sleep. I'm on disability, but I sleep instead of doing the things I enjoy.

@quietStorm247

9 күн бұрын

Wishing you better health. Please keep hoping.

@pogo55555

8 күн бұрын

Hang in there. My son is a survivor. It wasn't fun. A lot of people hear you. I wish you a full recovery. Try to enjoy some aspects of the downtime if you can. Good luck to you.

@PJ-hi1gz

8 күн бұрын

Wishing you the best, you’re a warrior.

@brexitgreens

8 күн бұрын

This is why we need less *AI safety* and more *AI progress.* Delay = genocide.

I use both Claude and ChatGPT. I pay for ChatGPT, but use the free version of Claude. I use them to help me with my job in education. I find Claude better for my specific tasks and questions.

@toadman5184

9 күн бұрын

Me too. I pay for both, plus CoPilot for Visual Studio. Claude can be very good with specific questions but tends to hallucinate quite quickly (coding).

@Neprow3000

8 күн бұрын

Is Claude has a memory? Or do you have to start from scratch with every interaction?

@JohnMcAfee-se9ms

8 күн бұрын

The students should just use the AI directly

@almightyzentaco

8 күн бұрын

Claude is far superior for bouncing complicated ideas against.

@HiddenPalm

8 күн бұрын

I'm the opposite. I pay for Claude and use gpt-4 for free through copilot.

EU: it should be a warning for all outside the EU, if companies do not support our regions. I read the act, and the act forces the companies to protect the user, especially misuse on private data. No support = data misuse. Think about.

Matt, I agree with you. GPT-4o couldn’t understand a simple question. I just now asked Claude the same question and after misunderstanding me it did eventually give me the correct response. I asked both to analyze an image I created in Leonardo and Claude gave me such a detailed reply that it made GPT seem short and rushed. Sadly, I quickly ran out of free credits for Claude lol.

@brexitgreens

8 күн бұрын

Try it via Poe AI.

@jerryoftheriver

21 сағат бұрын

Yup, its how they train the models, mostly to score on the hype pulpit.

7:38 Nakasone literally Just retired from the post of the Director of NSA in February 2024. 😮 You might as well just let NSA director be a board member directly. 😑😑

@soggybiscuit6098

8 күн бұрын

OpenAI might as well be skynet at this point in time

...idk man, I've been sending the same prompts to Chat and Claude simultaneously for we development, and Chatty 4o is still King.

OpenAI would be considered ahead by most people if they'd released the full 4o weeks ago (with the new voice mode and image generation)

@augustuslxiii

8 күн бұрын

I'm just questioning whether it's going to be in July or later, at this point. By all means, work out the problems before release, but clearly they should've said "in the coming months," at this rate.

@BionicAnimations

8 күн бұрын

Exactly. I am glad Anthropic came out with this update because maybe it will make OpenAI get off their butts and stop playing around with our emotions.

@robbievanwijk2512

8 күн бұрын

@@BionicAnimations Anthropic has good LLM's, but I feel like they haven't REALLY been competitions for OpenAI so far. Maybe this'll get them to get their act together.

OpenAI is going to have to change their name. Suggestions: GovAI, BigBrotherAI, SurveilAI, FireAllAI, NarcissAI.

@jge123

9 күн бұрын

Musk was right as always.

@SjarMenace

8 күн бұрын

Big brother ai hahahahahaha😂😂😂😂

@billfarley9015

7 күн бұрын

Skynet AI. Terminator AI.

@berkertaskiran

6 күн бұрын

My favorite is still ClosedAI.

@granand

6 күн бұрын

OpenAI is fine considering their POV was never you & me the users. Big Brothers knocks once the tech is in every gadget free and then thank you. Remember they were running Twitter and all meta products have open port to 3 letter agencies.

I have paid versions of all of the models and some via API. Things may look better on benchmarks, but in practical and useful output, 4o is still doing a better job for real world work.

@gumanelson2007

9 күн бұрын

No it isnt

@launchspeed9810

9 күн бұрын

Not for me either.

@ALFTHADRADDAD

9 күн бұрын

Disagree

@jason_v12345

9 күн бұрын

@@gumanelson2007 When you're working with models this close in ability, other factors are more at play: the specific task, the domain of knowledge, your own prompting skills....

@abonatlasanatate

9 күн бұрын

Until now, in my experience Opus 3 was better than Chat GPT 4o and 4. If Sonnet is better, OpenAI is in trouble

These benchmarks really need to go. They don’t give an indication of what these systems are like to use, and each iteration of the different models is barely better than the others. Also, have them write some real code, not just some stupid simplistic game. None of these models are able to make an iOS app that properly handles the camera in landscape mode.

@augustuslxiii

8 күн бұрын

The benchmarks can stay - I'd just like to actually see them played out instead of the Trust Me Bro results. (Apologies to Anthropic, et al, if they did in fact publish all that.)

@berkertaskiran

6 күн бұрын

Well Opus was utter crap since it even failed to say when you pass the 2nd you become the 2nd. And I haven't asked 3.5 this yet but it is impressive in such a way that I don't need to. I can already see it is much better than 4o. It is actually useful and makes recommendations that works instead of acting like a goddamn encyclopedia the 4o is. My take is when you're even 1-2% better on a few benchmarks you might be WAY better overall. It's just we need better benchmarks that actually shows the real differences. No other way to really tell except taking people's words for it or trying out which can always be misleading.

This over evaluation of every update made by LLMs is more of a problem than anything else. Most people just use them for the most mundane use. The most serious/important info given this week. Is the former NSA official. The gathering of information has reached a level, that most of these companies need to know how to SNEAK around the laws. To gather as much info as they can to strengthen their products. They eventually want the same rights Meta, KZread, Google, maybe we will not get swamped by ads, but nothing will be safe and I'm now wondering about owners' rights to thing discovered using AI? What will be the asking price? What will be the real cost to using A.I. But this over hyping...............

@computernerd8157

9 күн бұрын

True. I Say dont get addicted to this technology, if you do, you will become once again dependent on a company that will do Adobe and Unity things to you. For example, not owning any of your work or sooner or later all AI will become a services so without it alot of people over relying on it will become useless they have the cash.

@berkertaskiran

6 күн бұрын

Thing is, as AI rises, private data will reach to zero worth. There really won't be a use for it. AI will do all the work so why would anyone want anyone's data since no one does any job or anything meaningful? You don't see the real revolution.

I can’t believe I began watching this channel when it was at 50k…this is amazing to watch 📈

@jcriley7695

9 күн бұрын

It is insane isn't it. He deserves all the success as I'm sure he's on his way to make a killer living on it, if he isn't already.

@WillowsAIAcademy

9 күн бұрын

Did you watch when his videos were about gmail.

The problem with claude is you can't keep using it infinitely like chatgpt

@SjarMenace

8 күн бұрын

Even in paid version? Edit: it was a question, i forgot to put an question mark

@Ristaak

8 күн бұрын

Through Perplexity AI's API (which you get through there paid plan) you can use the Claude models indefinitely.

@brexitgreens

8 күн бұрын

@@Ristaak There's also the Poe AI route. With limited but generous tokens for $20pm.

@brexitgreens

8 күн бұрын

@@Ristaak Last but not least: you can deploy your own chat connected to any model on any API, including Claude, and paying only for what you use.

@nomadicgodking

8 күн бұрын

@@brexitgreensnow that’s the ticket right there. I’m bout to google….. I mean ChatGPT how to do it now!

OpenAI is falling behind except their total integrations with Microsoft, Apple, and the NSA.

@Redsword603

7 күн бұрын

Not only OpenAi is making themselves look bad, but also getting people raising suspicions about them for the latter.

@allanshpeley4284

7 күн бұрын

You guys really don't think OpenAI has far more advanced features that they're sitting on?

Great video, just one thing: Why nobody speak about how censored Claude is? Great for coding but almost useless for anything else due to the heavy auto censorship. And don't show him a woman in swimsuit he's going to burn a transistor... In comparison ChatGPT look like a degenerate (and it's saying a lot...).

The only way forward is through the kinds of opportunities that Olama allows. GPT and Claude are so painfully biased, censored, and deceptive. I understand the nature of bias in language models, but I'm talking about the additional overlay of content manipulation; it makes them borderline unusable. Whether motivated by data and attention mining or good intentions (the road to hell as they say), the myopia of the tech industry is, ironically, thwarting the potential of useful or interesting AI.

Matt reminds me of John Hammond from jurassic park, when he was excitedly staring at the baby raptor being hatched and mesmerizing at the birth without realizing what he's about to unleash 😢 🦕🤖

@poletooke4691

9 күн бұрын

If we're all gonna die, we may as well find and marvel at beauty and coolness in the world while we still can.

@brexitgreens

8 күн бұрын

​@@poletooke4691 A beautiful nemesis is much preferable to the continuing ugliness.

@OrgoneAlchemy

8 күн бұрын

The money and reputation of those steering frontier AI as well as big pharma and their robot pills and nanotech is very easily investigated. Only wilful ignorance and child like naivety can blind you to the transhumanist horrors being developed for the technocracy control grid. We can all be optimistic but it's foolish to do so at the expense of being realistic. Matt needs to wake the fuck up and delve a little more in to the dark side of big tech.

I have also found that GPT4o is worse than the older 4.0 I use it to create responses for customers. I will ask it to improve my note to the customer. The old GPT 4 used to understand the situation and craft a thoughtful note addresses what is needed. The new one basically takes my note and improves the grammar but that's about it. Old version was much smarter.

@augustuslxiii

8 күн бұрын

You can still use that version, you know.

I wouldn't call Europe's enhanced privacy protection 'unfortunate', especially not in the same video that expresses real fears of mass surveillance by AI. Hurrah for GDPR.

@mygirldarby

9 күн бұрын

He said they want the American AI companies to lower their security features.

@everyonecancraft70

9 күн бұрын

Problem with legislators is that they do not know what is really needed. They have to work with what in other spheres are called subject matter experts. The 2015 vat regulations were a case in point. EU tried to force some big fish to collect tax for them. In the process they forced hundreds of little fish out of business. When some little fish screamed we are here stop trying to kill us the EU said we didn't know you existed. They had tested their ideas with what they thought were small fish but who actually proved to be middle sized fish. In other words if the legislators are poorly advised by subject matter experts which seems to happen more frequently with tech than other things they cannot possibly make good laws.

@colinmorrison5119

9 күн бұрын

@@mygirldarby The Forbes article that covers Meta's reversal expressly mentioned GDPR.

@dhouchin83

8 күн бұрын

​@@mygirldarbyI must have missed something. How is that the case?

@mickmickymick6927

8 күн бұрын

The hypocrisy of US AI influencers is crazy. They shake their head at governments protecting their citizens, then they're like 'hmm, how are we going to protect our citizens from AI, if only there was some solution'

That internal pill story sounds like the movie Fantastic Voyage "1966" 😃

Sora from OpenAI waited so long to launch that it got surpassed by other technologies. With the release of Claude 3.5 Sonet by Anthropic, which offers superior performance and new features, it's clear that the competition has caught up. Meanwhile, Runway's Gen-3 and other video generation tools have also set new benchmarks. OpenAI needs to innovate quickly to stay relevant, as the AI landscape is evolving at a rapid pace.

Our T2 Remake was technically the first completely AI generated movie to feature in a theater (LA's landmark NuArt Theater) 50 of us came together to create the first feature length AI movie. Granted it was more like a long episode of robot Chicken it was still technically the first AI movie to be released and I was honored to be a part of the project

In the race of AI progress, even McDonald's couldn't ketchup - guess their drive-thru AI got fried before it could serve up any nuggets of wisdom!

We are close to AGI but we can't take a fast food order

@Arthur-jg4ji

8 күн бұрын

i think they cheap out on the ai side of thing or they failed to make their agent good

Claude is like a friend...so much nicer to use.....Other AIs are cold ❄️

I just asked Claude to translate a document in english cursive script from 1790. It nailed it!

Maybe unpopular opinion, but I wouldn't encourage OpenAI, or anyone else for that matter, to release something fast. I'd rather encourage them to take care of safety and security first. Considering recent discussions and rumors surrounding OpenAI and Sam, especially if they fully embrace a "for profit only" model, I'd be really concerned. As the large models become exponentially smarter, I wish they were becoming exponentially safer too.

The only reason why chatgpt seems to be "falling behind" is because most of their focus and power is going towards military/government projects now. If you think the government hasnt, or cant, seize control because "x, y, z", think again. They absolutely can do whatever they want, and controlling and using the top AI technology (that happens to be in their own country anyhow) will certainly interest them, and you can rest assured they already have their hands all over it. It's insane to think the american govnment isn't already in control.

@jason_v12345

9 күн бұрын

I think it's just a cat and mouse game. Company A releases their newest model, and everyone is amazed. Meanwhile, behind closed doors, company B is making unimaginable breakthroughs, until they release their next model, and everyone is amazed again, and the cycle repeats.

@RadelynFigueroa

9 күн бұрын

@@jason_v12345 Thats what it appears to be on the surface, yes. That is what we all expect. But what do you think the government/military is doing during all this? You think they are completely hands off and out of the game? I don't believe that for a second, and I think it has a great affect on everything going on, but it isn't talked about.

@jason_v12345

9 күн бұрын

@@RadelynFigueroa The running theory is that the military and OpenAI are in private partnership, and I'm sure if that were the case, that the contract would stipulate features that only the US military may access.

Either OpenAI has something really really big, thus the big name NSA dude. Or they are in trouble and bought this fella on to fix something. I always have said that AGI will come sooner than everyone thinks. I honestly think we will be very close by the end of this year. Timelines keep getting adjusted.... exponential growth is inevitable.

@somali4154

9 күн бұрын

I no longer trust OPENAI hype. Promise after promise without releasing the actual product.Competitors will soon overtake

@raul36

9 күн бұрын

Exponential growth is not possible for long periods of time mainly due to lack of resources. It is simply not possible. You can say what you want. AGI will not be there at the end of the year. By the way, what you are referring to is quadratic growth, not exponential. You guys tend to confuse it very easily.

@miauzure3960

9 күн бұрын

what are you talking about, we are orders of magnitude low on parameters of these models, to be as complex as human brains, and then there is lower complexity of synapses themselves. Biggest models are barely reliable in some easier day to day tasks, while their whole thinking and existing is text-based. To get even close to AGI you have to also have some kind of visual cortex and space orientation, or dynamic learning and synapse-reorienting, which is impossible with current technology.

@carlosamado7606

9 күн бұрын

​@@miauzure3960 Its not impossible with current tech, we just dont have the architecture for it. Neural spikes is being researched, capacity for memory, to learn indefinitely too. Robotics will probably aid on its functionality a lot too

@raul36

9 күн бұрын

@@carlosamado7606 You yourself have just clarified that the technology does not yet exist because it is being developed and researched on it. Do you even read what you write? 😂😂

The video 'AI Sentience in Art Gallery 2' video on KZread show why Mat's beardy thumbnail is so funny! Love it!

I swallowed a camera like that over 20 years ago under the supervision of Professor Polglaise...Gastroenterologist in Australia. it was new and experimental back then. Camera had transmitter and you had to wear a wire net receiving antenna across your abdomen. Afterwards they showed me the video of the journey of the camera....including the waterfall as I washed it down with water! Obviously something modern will be far more advanced and collect a lot more than just video. Thanks for your update.

@nathanbanks2354

9 күн бұрын

I think I'd rather collect the data wirelessly than eject a uSD card after or something....

Claude has worked better for me with coding for the last 2 months. Sonnet 3.5 is a huge step above gpt in my experience

Not Having Apple whitelabel GPT4 and putting shitty apps in their os isnt a loss for Europe.

One task Chat-GPT and other AI chat systems always fail at: using the Dominic Memory System, generate 100 celebrity names based on initials containing the letters, O, A, B, C, D, E, S, G, H, N. For each celebrity, assign a memorable action.

Also, OpenAI giving exclusive access to Sora to industry people really defeats the purpose of democratization of trades and even industrial revolutions. Like, you can't give crutches to get people through doing the same thing before and after a revolution. It defeats the purpose and prolongs the end result of new minds learning new tools creating new innovation. It really speaks volumes to Hollywoods incredible lack of faith in their own abilities. If you're good, AI will make you better. If you're not good, you're afraid of AI making other people better. I don't see how it is any more complicated than that.

With OpenAI just becoming just an arm of the Pentagon, I'll take any alternative.

@augustuslxiii

8 күн бұрын

Now, now, let's be fair. They're also an arm of Microsoft, which is also an arm of the Pentagon.

@brexitgreens

8 күн бұрын

​@@augustuslxiii With Anthropic being an arm of Amazon. And with Gemini being an arm of Google being an arm of Alphabet being an arm of the CIA. Our liberation can only come from Russia and China.

@allanshpeley4284

7 күн бұрын

I'll take whatever works best.

@brexitgreens

7 күн бұрын

@@augustuslxiii KZread deleted my comment referring to Google being an arm of the C▒A. 🤐

@brexitgreens

7 күн бұрын

censorship test 🤐

Another informative video…thx

Thanks for reporting!

Thanks Matt, great update as always!

Soon, social media users may start removing the metadata about AI. It's not difficult to remove. Perhaps, there is even a program for dat.

Rip Donald Sutherland @6:35

@quantumpotential7639

9 күн бұрын

A.i. will bring him back to life probably by next Tuesday. No one is going into the ground anymore. Just being put on ice till til A I. figures out how to resussitate them. I love Donald Sutherland. He was in the Dirty Dozen movie. He'll be back to star in more movies. Thanks

@berkertaskiran

6 күн бұрын

@@quantumpotential7639 He'll be back in movies but not in real life, sadly. Not even if AI makes a copy of him.

OpenAI seems to be all about the big announcements and flashy demos, but they're taking their sweet time getting products out to the public. Meanwhile, Anthropic just quietly drops Claude 3.5 Sonnet for everyone to use without much fuss. It really does come down to different company philosophies, doesn't it? OpenAI's under Sam Altman seems more focused on grabbing headlines, while Anthropic appears to prioritize getting their tech into people's hands. I wonder if this'll impact how successful each company is in the long run. What do you think?

I must be a weirdo because I am not that interested in chatbots. They waste so much of my time hallucinating, it's like paying money for a brain-damaged intern who lies a lot.

@rolestream

8 күн бұрын

Usually, it's a prompting issue. Check for issues between keyboard and chair.

@CozyChalet

8 күн бұрын

@@rolestream😂

@thymenwestrum7011

8 күн бұрын

Nah hallucinations are an unsolvable problem even with the best prompts​@@rolestream

@augustuslxiii

8 күн бұрын

If you don't use them, how can you attest to how often current models hallucinate?

@brexitgreens

8 күн бұрын

​@@augustuslxiii Possible explanation: she has used them before. It still creates a paradox though! Because, this way, she will eventually miss the point when her observation is no longer true. Also she's naïve to begin with by trusting anything or anyone other than chat bots. "Better the devil you know."

Also of note Claude 3.5's end of training date is April 2024 and it seems to be fairly detailed, I was asking about recent geopolitics and it was talking about the Red Sea shipping and incidents in Feb-April. I feel it was a bit more hallucinatory as you squash up against the April time-frame (probably because it doesn't have thousands of samples to reference in its data) and gets more accurate as you move backwards in time. Re: ChatGPT-4o I found it to be smart but stubborn. If it makes a mistake or you give it instructions it digs its heels in and won't obey or fix its misteps, Claude on the other hand is very chill, it will follow instructions well and will also tear apart its own answers if you ask it to check or redo.

We really appreciate your effort. Thank you for bringing the news to us.

Honestly, this channel is all you need to keep up with what's going on in AI

claude is so useless they still didnt release it in many countries. they dont even have built in voice mode. not sure how they plan to compete by showing these dumb benchmarks, when literally most people cant use it

So, now both SSI and chat gpt is under control of zoinist? and hiring people from USA state department sounds pretty bad to me.

This MidJourney text to video model has to slap hard now

What is that mini screen in the background? An old smart-phone or a special display?

I just want an ai that can make edits to a pdf. is that too much to ask?

@allanshpeley4284

7 күн бұрын

Why not just use Acrobat? Why do you need an AI for that?

@cosmic_pursuit

6 күн бұрын

@@allanshpeley4284 Ill try it - is it free?

Every company is saying much cheaper... much cheaper.... when are they gonna reduce the subscription fees then? I am an occasional user, I cant afford 20 USD per month.

@meow_meow_J

9 күн бұрын

Then use it via the API, I’m sure there are plenty of open source clients I’m using one right now, they scale based on usage, so if you truly are an occasional user, they shouldn’t be that expensive.

@viranijoush

9 күн бұрын

@@meow_meow_J how? I don't have much knowledge about using via api and clients etc.

With the new hire at OpenAI, keep in mind a lot of individuals rotate in and out of gov jobs, with those gov jobs specifically meant to deal with the companies they rotate in and out of. This isn't limited to tech companies, but they make up one of the largest groups.

I'd be all in on Anthropic if they had a setup that saved memory across all instances like GPT 4o - it is amazing on OS. I'll mention an idea and it'll complete the task without me even getting around to asking because it already knows everything about the brands. Crazy stuff.

Click bait. Deceptive title

At what point does calling yourself Open AI constitute false advertising?

Great info👍. Suno the music AI have big plans also😶‍🌫️

Hello !! Big fan, you didn’t talk about Hedra ?? Did you check it out ???

this video wont age well

@shadowproductions969

Күн бұрын

not really, it'll take a new ai model from openai to even compete

will AI get a seat in congress? real time fact checking would bring out the truth

Your video summaries are awesome. Thank you.

Thanks Matt…..love your comments & takes on the AI news

I hate so much click bate technics!

Worst clickbait title. OMG Predict some pushback here

Open AI might release GPT5, but it'd be JUST GREAT if they could finish releasing 4.O.

Great video as always but no chapters this time?

Human Intelligence=A bunch of people wanting to be the first to comment by writting the intelligent phrase “first”. Good luck humanity🌍🤔

@pbft.j

9 күн бұрын

It's a game. Actually quite a big reason for human innovation. Humans try. Humans enjoy competition. There's something enjoyable about the mundane. You, however, don't need to worry about such things. Elizabeth is quite the effin' smart ass huh?

@user-ty9ho4ct4k

9 күн бұрын

Yes. The elephant in the room is that humans aren't that intelligent.

Lovely clickbait 🪠

Honest to Goodness, You are Fabulous!!! I love your enthusiasm, your personal thoughts & stories, & you tell us MEANINGFUL stuff totally delightfully!! I am grateful for you! Thank You! 🎉🙏🌟🎁❤

What concerns me about the delay in releasing these incredible new image generators is that their best features may be bought up, or their developers bought out, by the major media companies. We could be left with access to the most basic offerings (incredible as those will be) while advancements will be in the hands of giant corporations. Hope I’m wrong.

I’m really dismayed that you chose a clickbait title. I expect better from you

Wait!!! You're trusting benchmarks FROM the company that created Claude 3.5 Sonnet? Do you see what the problem might be there? Honestly, that's just stupid and amateurish. Do better.

Great insights, Matt! Your detailed breakdown of the AI landscape, from Claude's coding prowess to the implications of OpenAI's collaborations, really helps us navigate this rapidly evolving tech world. Keep up the fantastic work!

As I see this benchmarks the Sonet is in practice on the same level as GPT-4o and is beating it in the margin of error so claims that it is new king is a bit overstatement

Lol... people complaining about the thumbnail and click bait 😂 Those comments are getting so old... I came here because it's Matt's channel and one of the best resources for AI news. Also, people who are Anti-A.I. are like Luke Skywalker, getting whole planets wiped out with dumb things like 'hope', as their motivation, when joining the Empire would be way better... better uniforms, transport, accommodation and probably better medical too...😂😂😂😂 Thanks for another great vid. Appreciated. 🙌

Great episode. it was all just a matter of time. Thanks for sharing!

Thanks Matt great content🎉. Totally agree about ChatGPT ‘dumbing down’ that’s my experience too. Also haven’t received theChatGPT 4o updates. Although I am in the Uk so that might be why😢, Keep up the fantastic work work!!👍🏽

@VinMan-ql1yu

9 күн бұрын

same, gpt4 was great then became good. 4o is OK, faster, cheaper, but not smarter

hmm, it's not on Chatbot Arena yet. I wonder how long it takes to show up there.

@matt, where can I get the digital metric board you have in the background? would love to have one by myself for my company

At the moment Claude 3.5 is my default model - but I use as well GPT4 and Gemini 1.5 Pro - for specific tasks.

This is very insightful. You obviously put a lot of time into your videos. Thank you 😊

This is such an amazing space, awesome watch I only understand 20% though learning well, what a time we live in! 🎉

Claude is great at modern front-end development. For writing Next.js 14 API routes with me yesterday, it was correct the first time about 50%, and got there after a few error feedback loops for the rest. No editing from me. I did seven API routes that used AI API calls. This wasn't something trivial. I bought the pro account immediately.

I think this fierce competition between companies is amazing, just look how fast AI videos and language models are advancing.

The product vs feature that Marcus aired explains this nicely

Very helpful- thanks

8:15 honestly... LLM is perfect for language interpolation, and conducting unofficial automated audits of everyone's messaging. I always thought this from the first time it was being developed back in college. It was perfect for understanding You mean language throughout different cultures and dialects