Claude 3.5 is mindblowing at translation too. Try it in your own language. In Bulgarian it killed it. Not like the way it was before. That's truly impressive!
I can confirm. Claude 3.5 Sonnet can _almost_ do rhymed poetry in other languages. It can speak expertly in _antique variants_ of the said languages. Meanwhile GPT-4o stumbles in their basic form.
@@brexitgreens نعم صحيح كلود 3.5 يكتب قصائد عربية قديمة صحيحة الوزن والقافية وهي صعبة و تستخدم لغة قديمة تقليدية ذات جزالة وقوة ومتانة في اللغة ، كما أن gpt4o لا يستطيع كتابة بيت واحدة صحيح الوزن هههه😊
I've been on the fence for the last few weeks about canceling my OpenAI subscription and getting Claude instead, I think it's time to support a company that actually brings us new and exciting things and not just brags about things they do internally without giving us access
Anthropic needs to release web capabilities ASAP and its own form of Assistants and it will be extremely competitive. I am excited to see what they come up with! Thanks for this video!
@@JhonataCosmo Yep! And image generation and their own GPTs. However, they also need to ease up the censorship. They do all these things, and I am switching from OpenAi to Anthropic in a quick second.
@@qwax- We need to quit confusing safety with naughty words and political incorrectness, and such. Agents that go and hack the systems at a fission plant in response to some request -- that's a safety issue.
There's so much room for improvement in AI UX. We are only just coming out of the command line interface stage. Like going from DOS to Windows 1.0 One obvious next add is a drop down list of system prompts in the developer console, which can be updated from other users as well as the prompt library so it's instantly accessible as well as being set by other agents. So you could have a manager agent create a new task and worker agent using just the existing name/id of the system prompt from the list. As a developer I think of the actor model of programming that has agents with their own threads passing messages between them. This was created way back in the 70's when SmallTalk was the language and agents were simple programs but that got overshadowed by object orientation which seemed simpler but now with intelligent agents I can see some of that theory coming back with prototypes being AI agents passing messages and then being optimised over time. We're just getting started.
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, The unique voice feature of OpenAI sets it apart from all other AI systems. The ability for seamless, real-time interaction, where you can interrupt the AI and it picks up right where it left off, creates a level of intimacy and natural communication unmatched by others. OpenAI's ChatGPT-4 Omni can laugh, pretend to cry, and even modulate its voice features that no other AI currently offers. The new vision capability, allowing for back-and-forth conversation while visually interacting with users, is another groundbreaking feature exclusive to OpenAI. This combination of advanced voice and vision capabilities makes OpenAI the leader in AI technology, setting it apart from the competition. This is why OpenAI is regarded as the best it not only excels in performance but also provides an interactive experience that no other AI can match. It's clear that to compete with OpenAI, others must adopt these innovative feature, but they can’t.
I take it you are using Apple Silicon. For the rest of us the voice features do not work. That sucks when you are paying a monthly fee. As soon as Anthropic increase their prompt limit, I know where my money is going.
I've meet perfectionists that won't ship anything until is perfect. I have also met very rich people that barely knew how to code, but they could sell an OK version till a better version was ready. Weird things at Open AI with MS and Apple both leaving. Legal department was probably afraid of what's coming, whatever that is.
Ends of 2025 equals eternity in the AI fields, and PhD level is a freaking high bar to meet, if they don't deliver they will be done as a company, but if they do, the job market won't never be the same again.
Wes Roth posted a video recently where he showed that openai has such a huge lead over its competition in terms of world internet traffic and chatgpt name recognition that they have no need to be concerned about what the competition is doing. They are busy securing contracts and deals with every major player that they can including the government.
Yes, it is true, Claude 3.5 writes ancient Arabic poems with correct meter and rhyme, which are difficult and uses an ancient, traditional language that has magnificence, strength, and durability in the language. Also, gpt4o cannot write a single verse with correct meter, haha.
Check out cody from sourcegraph vs code extension. Unlimited prompts except if you do like 20 prompts per minute in which case they might ban. I use cody in my 8 hour job and I have never been limited at all.
Yes, I'm not happy paying OpenAI to develop things for environments I don't have (Apple Silicon) and teasing but not releasing new features. I'd switch from OpenAI but the prompt limit for a *paid* account is a deal breaker.
Both GPT-5 and Claude 4 will have AGI levels of intelligence. OpenAI will probably release it a few months earlier, as they are ahead in development, and it will definitely not be as late as Q4 of 2025, at worst early summer of 2025 if not earlier (I expect Q1 of 2025 for GPT-5, unless something major happen from someone else, and they release it even faster). Also, both will be irrelevant by the end of 2027 when we will be able to run AGI-level models (open source and uncensored) on consumer hardware which even if expensive at the start, it will be at worst around the same price as an RTX 4090, but 100s if not 1000s of times better for running and training such models. For now I am taking advantage of both GPT-4o and Claude 3.5 Sonnet as they are free (even if limited as such), and they are both great for brainstorming and expanding ideas for my TTRPG and the campaigns I run with my system (game design, abilities, lore, quests, NPCs, etc.)
Sora has been released to “red teams”. In the name of responsibility, gpt5(or whatever it will be called) won’t be released until after the US election.
They cherry picked and fakes the Sora videos and the voice model was meh at best. They don't release it to the public because it's not going to be good enough to keep the illusion going.
@@YeeLeeHaw yeah but the sora reveal was quite awhile ago, there's no reason to believe they haven't been non-stop working on that model since. they'll keep improving it until it actually publicly releases.
@@georgelong7153 I think OpenAI just realized that being involved in a scandal where fake videos generated in Sora had some kind of effect on the election would have been really bad for their image and public acceptance of their products. Letting other companies test those waters while they improve theirs in the background is a smart approach, especially if they also don't really want to be the company leading massive advancements in AI tech, so they're naturally waiting for other companies to get close to them before releasing the next updates.
Sam Altman has done a few interviews, one with Lex Fridman, where they talked about disruptive technology and having to reduce the "shock factor" (I'm paraphrasing) What they're doing is pretty much on-par with what he said. If you let the public know about it, and cook on it a while, then they want it. If you just release it, there's tons of backlash if it's a disruptive technology.
That can be true .. Or can be what we wanna hear to keep the hype going... I'll believe they have something disruptive when I see proof of this. Until then, we cannot forget this is a hype business and they need to sell it
I've used gpt4 to his effect, but lately moved 90% to Claude sonnet. It's simply easier to get what i need done. It understands me better with less effort.
I find that Claude and ChatGPT each have their own pros and cons, that currently justify me still using both for different things. Claude is ideal for coding and writing and seems to have a better comprehension of complicated concepts, be it mathematic, science, or creative. But the chat limit is too small, even with a premium account, for long sessions. Chat GPT's voice mode, even without the new upcoming update, is just so much better for any casual conversations or to role play or have a back and forth verbal conversation to explore ideas. Speaking Hands free is absolutely under-rated. Plus, the GPT Store and Custom GPTs feel a bit more intuitive to make and use, and currently offer more variety than Claude's Projects. It can also make images, has less aggressive censorship, and can hold long conversations without hitting a message limit. Chat GPT also seems to have a bit more of a fun personalit, by default, where Claude is excellent at embodying a more technical or tone. And, although Claude us technically skater, GPT is still no slouch.
These LLMs still have major issues with hallucinations and following instructions. So, I can see OpenAI delaying GPT-5 because they may have reached the limits of the transformer model. If GPT-5 gets released and it is just marginally better than GPT-4, I think the whole AI hype and GPU explosion collapses bringing down the magnificent 7 and the S&P500. As long as we are anticipating GPT-5 and AGI, the money will continue to pour in.
The problem with Anthropic products is even heavier censorship. As if it wasn’t enough with ChatGPT already… Claude doesn’t want to discuss any sensitive topics at all. People post lots of disturbing answers from it in internet, where they ask the model about WW2 events, etc, and the model just rejects, saying it’s too sensitive. It’s ridiculous!
the thing is that reguardless of how good Claude is OpenAI hold the VAST majority of the AI market with ChatGPT and 95% of those users arent super AI enthusiasts that will flock to the technically best new thing ChatGPT still is good so there's really no reason to and OpenAI knows this which is why they don't really care to ship it doesn't effect most of their Users trust only the AI enthusiasts
People should be able to find the best thing available, I don't consider myself an AI enthusiast, but if I see that a service is not the best service out there, why am I spending my money in it ?
@@Pawnsappsee I absolutely consider myself an AI enthusiast and I pay for both Claude and ChatGPT because they each have plenty of benefits but when it comes down to it I still prefer ChatGPt most of the time so I don't really get the Claude hype and people trashing on ChatGPT as if its all of a sudden dumb or something
We can all hate on OpenAI for giving us nothing, but when they actually release their next model, it will blow previous models out of the water guaranteed, same as GPT-4
In the two primary applications I use LLMs for, coding and writing, Claude 3.5-sonner is superior to GPT-4o. Also, I'm scared of and distrust OpenAI as a company.
I would love someone would give you a programming crash course even if AI will remplace that in few years you deserve the best for yourself and... you would not need to be an expert just a basic coding skills you would be proud of using to test the mighty LLMs an AIAs in any case you are getting above many other UA-camrs in the field from my personal perspective you are getting better over time even if you where already amazing to start with!!!
Hmmm. This is evidence of the opposite for me. OpenAI is so confident in its position, it doesn't need to release every 5 seconds. Its aggressive releases in the past achieved clear market dominance and now they can relax a little with the release schedule while still remaining at substantially higher market penetration than any of its competitors. Also, as models get more advanced, they will take longer to reach consumers because more red teaming and testing is needed to ensure their safety and minimize their abusability. As models get closer to where ChatGPT 5 undoubtedly already is, they will see the same restrictions and delays and government intervention that OpenAI currently is with its frontier models.
I'm so annoyed that Sora hasn't been released after the teasing. I wonder if Claude will be working on their own AI video creator. If so and it's not a "dangle" I'll be a full-time supporter, never to look back at ChatGPT again after how they are only allowing large companies access instead of the public.
My test for a model is just having a conversation with it about interesting hypotheticals. Code or otherwise. My measure of success if how well it follows along, or the ideas it returns.
I don't understand why people praise anthropic. Their Claude AI is pretty much the most censored AI ever. It feels like I am being lectured everytime I deviate from their view.
@@2AoDqqLTU5v I had to work with statistics from the police and Claude would not let me analyze them because it was lecturing me how I had to factor in their background and race...
OpenAI knows that the underlying model structure for its behind the curtain unreleased model hit its limits so instead of focussing on that and essentially taking a potato and turning it into fries and then chips and then mashed potatoes. Instead of that, they are learning a whole new fundamental structure that can be truly scaled. They are thinking long term, no longer milking the short term incremental releases like they were originally.
Matt need more in depth videos about claude. I say to everybody that claude is better than open AI but nobody just listens. everything just know about that old chatgpt. But claude is awesome chatgpt is nothing in front of claude. The main point claude lets us test and make and do stuff. Open AI just shocases stuff to us consumers.
Wait to be disappointed in OAI until the election is over. If they dont release some sugar then, I too will be disappointed and disillusioned with them. Giving them the benefit of the doubt.
OpenAI is all but owned by the US government at this point with a heavy dose of dependency on Microsoft so it’s very clear how it’s being used internally by the track record of simply those two influences. 😂😢
Perplexity is actually kicking ass right now. Ive been using it a lot as of late when i need intenet searchib vs open ai. Not sure if its a wrapper to something
ive been wanting an extension to make it so I watch youtube less, and I originally tried with GPT 4 and it was just a lot of back and forth, nearly every output would have a bug or not work. When I tried with Claude 3.5 its only encountered an issue like once. And of course made the basic idea in a single oneshot prompt.
It’s certainly doesn’t feel like we’re on an exponential growth trajectory. I think diffusion models have basically reached their peak and LLMs may not be far behind.
I mean, LLMS have reached their peak, Image generation has reached its peak, hell, even video generations have already almost reached its peak! I think now we have mastered everything, and we can start working on AI robots.
Openai; NSA board member, forced to remove Apple and Microsoft board members, pulled out of China and now 'partnering' with Los Alamos by executive order. Openai is in governmental special measures.
OpenAI is just moving slower like an investor. While others AI companies are just gathering coins, OpenAI instead is building HUGE things. That’s why they don’t care about 20$ subscribers. They are taking care of real deals out there. Take a note of my comment.
i'll bet you that openAI sora was hype to boost its market value. But considering how its inevitable that competing or open source versions are going happen- Why would any company contract with sora? Paid access would be a mistake, unless you could profit from using it right now. Iunno tho... lots of questions...
We said that about dall e 2, when it was released to the public guess what, midjourney came and swept dall e off their feet! Same with dall e 3, midjourney was already better, but then stable diffusion 3 came and swept dall e 3 off it's feet! Sora is by far the most closed off model open ai has announced, and guess what.... Sora is already almost getting swept off it's feet! If open ai wants to stay on top, they have to speed things up.
Inferior products often destroy better products in the marketplace - actually, it is a very common when the majority of a product's users are not capable of understanding how much a product is capable of
OpenAI faked both SORA and the advanced voice models. It's so obvious to anybody not caught up in the hype. Any other company would let people have access to the tools to keep the hype going. Or they'd at least keep releasing more examples on their own every week. Nope. Nothing. Crickets. They're busy trying to figure out how to escape the inevitable PR nightmare when the truth comes out.
Select people have used SORA. It's obviously not fake. More likely they have compute capacity/infrastructure problems to sort out. If they opened it up for everyone tomorrow, it may well be unacceptably slow.
@@kevnarTbf, you're claiming something that has absolutely no evidence based on what you "Feel" other companies would do with the same tech-- If you had some kind of substantial evidence it'd be a bit different. Also some have Used Sora.
I got tired of getting woke shamed and bullied by openAI all day long so I’m using Claude. It’s woke too but way more toned back and more acceptable levels of censorship.
ohh have you not seen all these advertised for download on facebook.. i would not risk downloading they are defo fake though using all the stock footage and making themselves out like they wrote the software etc.. and name looks close but are not the creators
![Flux.1 takes AI Image Customization to a whole new Level. [SUPER EASY GUIDE]](http://i.ytimg.com/vi/FvCgYQgQexo/mqdefault.jpg)
![АЛАУДИНОВ у Скабеевой: эти их ВСУ нас НЕ ДОГОНЯТ 😁 [Пародия]](http://i.ytimg.com/vi/an0anqU4WGQ/mqdefault.jpg)
Disappointed in OpenAIs lack of new features. I switched to Claude for the time being for coding
I am doing the same thing.
Switched to Claude a few months ago, but will keep an eye open on OpenAI
I thougt claude ai is much better than gpt 4o. Its absolutely fu**ing not. Its trash! Gpt4o understands my prompts better and gives better results!
@Entity303GB Why are you giving a chatbot prompts?
@@Entity303GB- Fake comment!
Claude 3.5 is mindblowing at translation too. Try it in your own language. In Bulgarian it killed it. Not like the way it was before. That's truly impressive!
I can confirm. Claude 3.5 Sonnet can _almost_ do rhymed poetry in other languages. It can speak expertly in _antique variants_ of the said languages. Meanwhile GPT-4o stumbles in their basic form.
@@brexitgreens
نعم صحيح كلود 3.5 يكتب قصائد عربية قديمة صحيحة الوزن والقافية وهي صعبة و تستخدم لغة قديمة تقليدية ذات جزالة وقوة ومتانة في اللغة ، كما أن gpt4o لا يستطيع كتابة بيت واحدة صحيح الوزن هههه😊
@@ilyass-alami Impressive model!
4o is also the best openai model for translation (one of the rare things its good at) so will have to test 3.5 sonnet for sure
I've been on the fence for the last few weeks about canceling my OpenAI subscription and getting Claude instead, I think it's time to support a company that actually brings us new and exciting things and not just brags about things they do internally without giving us access
OpenAI now works with the big guys. Note the recent board add.
I.E. United States government
That didn’t end up happening
Sam Altman alone should tell us. Venture capital background. He's in it for profit first and foremost.
Recent board adds were reversed uno
Lol yeah it's all about money bruh
I fear there are too many politics within Open AI......
There has to be. Similarly how we all came to an agreement not to nuke the fuck out of eachother
What does that mean?
Why? You planning to go work for them?
@@FrotLopOfficial yes it should follow the LAW... nothing else.
Anthropic needs to release web capabilities ASAP and its own form of Assistants and it will be extremely competitive. I am excited to see what they come up with! Thanks for this video!
Agents needs to be their next move
Agreed.👍
@@JhonataCosmo Yep! And image generation and their own GPTs. However, they also need to ease up the censorship. They do all these things, and I am switching from OpenAi to Anthropic in a quick second.
@@BionicAnimations They aren't going to do that. They believe heavily in safety,
@@qwax- We need to quit confusing safety with naughty words and political incorrectness, and such. Agents that go and hack the systems at a fission plant in response to some request -- that's a safety issue.
There's so much room for improvement in AI UX.
We are only just coming out of the command line interface stage. Like going from DOS to Windows 1.0
One obvious next add is a drop down list of system prompts in the developer console, which can be updated from other users as well as the prompt library so it's instantly accessible as well as being set by other agents. So you could have a manager agent create a new task and worker agent using just the existing name/id of the system prompt from the list.
As a developer I think of the actor model of programming that has agents with their own threads passing messages between them. This was created way back in the 70's when SmallTalk was the language and agents were simple programs but that got overshadowed by object orientation which seemed simpler but now with intelligent agents I can see some of that theory coming back with prototypes being AI agents passing messages and then being optimised over time.
We're just getting started.
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, The unique voice feature of OpenAI sets it apart from all other AI systems. The ability for seamless, real-time interaction, where you can interrupt the AI and it picks up right where it left off, creates a level of intimacy and natural communication unmatched by others. OpenAI's ChatGPT-4 Omni can laugh, pretend to cry, and even modulate its voice features that no other AI currently offers. The new vision capability, allowing for back-and-forth conversation while visually interacting with users, is another groundbreaking feature exclusive to OpenAI. This combination of advanced voice and vision capabilities makes OpenAI the leader in AI technology, setting it apart from the competition. This is why OpenAI is regarded as the best it not only excels in performance but also provides an interactive experience that no other AI can match. It's clear that to compete with OpenAI, others must adopt these innovative feature, but they can’t.
Or you’re an AI or you have to learn how to synthesize texts
I take it you are using Apple Silicon. For the rest of us the voice features do not work. That sucks when you are paying a monthly fee. As soon as Anthropic increase their prompt limit, I know where my money is going.
Anthropic is better.
I've meet perfectionists that won't ship anything until is perfect. I have also met very rich people that barely knew how to code, but they could sell an OK version till a better version was ready. Weird things at Open AI with MS and Apple both leaving. Legal department was probably afraid of what's coming, whatever that is.
Ends of 2025 equals eternity in the AI fields, and PhD level is a freaking high bar to meet, if they don't deliver they will be done as a company, but if they do, the job market won't never be the same again.
It is unique that Apple and Microsoft actually agree on something. But yeah, keen to understand why
Cool, but let me know when Anthropic ships a functional text box. Standard navigation keys simply do not work, such as HOME and END.
The cute crab game reminds me how excited
I was about Pong.
Wes Roth posted a video recently where he showed that openai has such a huge lead over its competition in terms of world internet traffic and chatgpt name recognition that they have no need to be concerned about what the competition is doing. They are busy securing contracts and deals with every major player that they can including the government.
Reaping the fruit instead of sowing. Won't end well for them. Unless becoming a government agency counts as "success".
Apple was ahead too… with the first iPhone .. I’d say they were 10 yrs ahead.
Which government?
@@brexitgreens if Raytheon and Lockheed aren't departments then we don't have to worry about oepnai ever being one. Not with Altman as ceo anyway.
that clickbait failure?
Yes, it is true, Claude 3.5 writes ancient Arabic poems with correct meter and rhyme, which are difficult and uses an ancient, traditional language that has magnificence, strength, and durability in the language. Also, gpt4o cannot write a single verse with correct meter, haha.
Great example that I wouldn't have tried personally. Cool!
25 prompts per day for over $20 per month is unusable.
Check out cody from sourcegraph vs code extension. Unlimited prompts except if you do like 20 prompts per minute in which case they might ban. I use cody in my 8 hour job and I have never been limited at all.
Indeed!
Yeah that sucks, but simple solution is API access and chatbot-ui
Yes, I'm not happy paying OpenAI to develop things for environments I don't have (Apple Silicon) and teasing but not releasing new features. I'd switch from OpenAI but the prompt limit for a *paid* account is a deal breaker.
bruh what? you can do that even with free
Both GPT-5 and Claude 4 will have AGI levels of intelligence.
OpenAI will probably release it a few months earlier, as they are ahead in development, and it will definitely not be as late as Q4 of 2025, at worst early summer of 2025 if not earlier (I expect Q1 of 2025 for GPT-5, unless something major happen from someone else, and they release it even faster).
Also, both will be irrelevant by the end of 2027 when we will be able to run AGI-level models (open source and uncensored) on consumer hardware which even if expensive at the start, it will be at worst around the same price as an RTX 4090, but 100s if not 1000s of times better for running and training such models.
For now I am taking advantage of both GPT-4o and Claude 3.5 Sonnet as they are free (even if limited as such), and they are both great for brainstorming and expanding ideas for my TTRPG and the campaigns I run with my system (game design, abilities, lore, quests, NPCs, etc.)
This version of claude is definitely changing the world, this is a huge step to wherever we are going
Sora has been released to “red teams”. In the name of responsibility, gpt5(or whatever it will be called) won’t be released until after the US election.
They cherry picked and fakes the Sora videos and the voice model was meh at best. They don't release it to the public because it's not going to be good enough to keep the illusion going.
@@YeeLeeHaw yeah but the sora reveal was quite awhile ago, there's no reason to believe they haven't been non-stop working on that model since. they'll keep improving it until it actually publicly releases.
Oh thank God they saved the sanctity of our sublime democracy. Thank you Open AI for truly caring about the citizens! 😅😅
@@georgelong7153 I think OpenAI just realized that being involved in a scandal where fake videos generated in Sora had some kind of effect on the election would have been really bad for their image and public acceptance of their products. Letting other companies test those waters while they improve theirs in the background is a smart approach, especially if they also don't really want to be the company leading massive advancements in AI tech, so they're naturally waiting for other companies to get close to them before releasing the next updates.
Sam Altman has done a few interviews, one with Lex Fridman, where they talked about disruptive technology and having to reduce the "shock factor" (I'm paraphrasing) What they're doing is pretty much on-par with what he said. If you let the public know about it, and cook on it a while, then they want it. If you just release it, there's tons of backlash if it's a disruptive technology.
Yes
That can be true ..
Or can be what we wanna hear to keep the hype going...
I'll believe they have something disruptive when I see proof of this.
Until then, we cannot forget this is a hype business and they need to sell it
I love your test examples! Great mix of quality info and entertainment. You rock, Matt!
I've used gpt4 to his effect, but lately moved 90% to Claude sonnet. It's simply easier to get what i need done. It understands me better with less effort.
Claude needs voice. I'm honestly mad at openAI the way they teased their voice and said it would be released in the coming weeks.
Pouring one out for the poor Frog 😢 😂
I find that Claude and ChatGPT each have their own pros and cons, that currently justify me still using both for different things.
Claude is ideal for coding and writing and seems to have a better comprehension of complicated concepts, be it mathematic, science, or creative. But the chat limit is too small, even with a premium account, for long sessions.
Chat GPT's voice mode, even without the new upcoming update, is just so much better for any casual conversations or to role play or have a back and forth verbal conversation to explore ideas. Speaking Hands free is absolutely under-rated. Plus, the GPT Store and Custom GPTs feel a bit more intuitive to make and use, and currently offer more variety than Claude's Projects. It can also make images, has less aggressive censorship, and can hold long conversations without hitting a message limit.
Chat GPT also seems to have a bit more of a fun personalit, by default, where Claude is excellent at embodying a more technical or tone. And, although Claude us technically skater, GPT is still no slouch.
These LLMs still have major issues with hallucinations and following instructions. So, I can see OpenAI delaying GPT-5 because they may have reached the limits of the transformer model. If GPT-5 gets released and it is just marginally better than GPT-4, I think the whole AI hype and GPU explosion collapses bringing down the magnificent 7 and the S&P500. As long as we are anticipating GPT-5 and AGI, the money will continue to pour in.
I didn’t know about the artifact sharing thing. Seems like someday the world’s most extensive video game is going to be open source.
I fkn love your benchmark testing approach!😂
The problem with Anthropic products is even heavier censorship. As if it wasn’t enough with ChatGPT already… Claude doesn’t want to discuss any sensitive topics at all. People post lots of disturbing answers from it in internet, where they ask the model about WW2 events, etc, and the model just rejects, saying it’s too sensitive. It’s ridiculous!
That "variables" stuff is the beginning of object-oriented LLM inference...
Im excited to see opus.
Opus has been out. Used it several times
@@sanchez-C137 I mean Opus 3.5
Imagine being OpenAI making videos on how their new video chat will help blind people.. *and then delay it this long blue-balling them.*
only their grandchildren desreve it, not you, unless you are a multi billionaire
Great overview and catchup, always appreciate your work!
I'd probably switch over to Claud 100% if it had Web capabilities. Might combo alongside Perplexity rather than OpenAi... For now
And less censorship.
Also how do you get it to actually show you artifacts? When I try this it tells me that it currently cannot preview code?
the thing is that reguardless of how good Claude is OpenAI hold the VAST majority of the AI market with ChatGPT and 95% of those users arent super AI enthusiasts that will flock to the technically best new thing ChatGPT still is good so there's really no reason to and OpenAI knows this which is why they don't really care to ship it doesn't effect most of their Users trust only the AI enthusiasts
People should be able to find the best thing available, I don't consider myself an AI enthusiast, but if I see that a service is not the best service out there, why am I spending my money in it ?
@@Pawnsappsee I absolutely consider myself an AI enthusiast and I pay for both Claude and ChatGPT because they each have plenty of benefits but when it comes down to it I still prefer ChatGPt most of the time so I don't really get the Claude hype and people trashing on ChatGPT as if its all of a sudden dumb or something
ChatGPT is still the most feature rich platform, nothing else comes close. Actually LLM power and quality is another thing entirely.
We can all hate on OpenAI for giving us nothing, but when they actually release their next model, it will blow previous models out of the water guaranteed, same as GPT-4
Sonnet 3.5 is notably better than GPT-4o. It's not as polished, but it has deeper insights and better expert-level knowledge.
In the two primary applications I use LLMs for, coding and writing, Claude 3.5-sonner is superior to GPT-4o. Also, I'm scared of and distrust OpenAI as a company.
I would love someone would give you a programming crash course even if AI will remplace that in few years you deserve the best for yourself and... you would not need to be an expert just a basic coding skills you would be proud of using to test the mighty LLMs an AIAs in any case you are getting above many other UA-camrs in the field from my personal perspective you are getting better over time even if you where already amazing to start with!!!
Hmmm. This is evidence of the opposite for me. OpenAI is so confident in its position, it doesn't need to release every 5 seconds. Its aggressive releases in the past achieved clear market dominance and now they can relax a little with the release schedule while still remaining at substantially higher market penetration than any of its competitors. Also, as models get more advanced, they will take longer to reach consumers because more red teaming and testing is needed to ensure their safety and minimize their abusability. As models get closer to where ChatGPT 5 undoubtedly already is, they will see the same restrictions and delays and government intervention that OpenAI currently is with its frontier models.
14:20 - can the simulation itself change the value of the variables? That would be cool.
I'm so annoyed that Sora hasn't been released after the teasing. I wonder if Claude will be working on their own AI video creator. If so and it's not a "dangle" I'll be a full-time supporter, never to look back at ChatGPT again after how they are only allowing large companies access instead of the public.
I think Open A.I needs to be the MySpace of this era and just vanish, there is selling out but this is on a whole other level.
My test for a model is just having a conversation with it about interesting hypotheticals. Code or otherwise. My measure of success if how well it follows along, or the ideas it returns.
I don't understand why people praise anthropic. Their Claude AI is pretty much the most censored AI ever. It feels like I am being lectured everytime I deviate from their view.
Exactly. And then it starts yapping about how it's bad. Atleast open ai gives a simple answer: "Sorry, but I can't assist with that."
I found that both chatbots were extraordinarily neutral when i used them during the european elections.
You clearly don’t really use AI for work lol. Claude is 100% better.
@@2AoDqqLTU5v I had to work with statistics from the police and Claude would not let me analyze them because it was lecturing me how I had to factor in their background and race...
3.5 is better than whatver claude 2 was but i believe 4o is still more lenient than 3.5 but i have not done extensive testing
OpenAI knows that the underlying model structure for its behind the curtain unreleased model hit its limits so instead of focussing on that and essentially taking a potato and turning it into fries and then chips and then mashed potatoes. Instead of that, they are learning a whole new fundamental structure that can be truly scaled. They are thinking long term, no longer milking the short term incremental releases like they were originally.
Matt need more in depth videos about claude. I say to everybody that claude is better than open AI but nobody just listens. everything just know about that old chatgpt. But claude is awesome chatgpt is nothing in front of claude. The main point claude lets us test and make and do stuff. Open AI just shocases stuff to us consumers.
Wait to be disappointed in OAI until the election is over. If they dont release some sugar then, I too will be disappointed and disillusioned with them. Giving them the benefit of the doubt.
OpenAI is all but owned by the US government at this point with a heavy dose of dependency on Microsoft so it’s very clear how it’s being used internally by the track record of simply those two influences. 😂😢
How? Because Microsoft? I don't understand 😢
They are not in bed with the Gov
I wonder if open AI wants to wait until the US elections are over, before releasing sora depending on how powerful it is could be problematic
This is the real world when some things just taking time and it's not necessary for every company to introduce new amazing product every week
No fun for the masses actually means no power to the people!
Maybe for the better? But I still think that it’s not up to them to decide…
Perplexity is actually kicking ass right now. Ive been using it a lot as of late when i need intenet searchib vs open ai. Not sure if its a wrapper to something
I think that open ai is working on something bigger, more fundamentally different, so i would not count them out yet.
ive been wanting an extension to make it so I watch youtube less, and I originally tried with GPT 4 and it was just a lot of back and forth, nearly every output would have a bug or not work.
When I tried with Claude 3.5 its only encountered an issue like once. And of course made the basic idea in a single oneshot prompt.
Claudes the best hands down
does anyone else feel like nothing happened in ai tha last 6 months?
It’s certainly doesn’t feel like we’re on an exponential growth trajectory. I think diffusion models have basically reached their peak and LLMs may not be far behind.
I blame the collapse of Stability AI.
@@lamsmiley1944 diffusion model searchers just choosed the safety instead of the growth
I mean, LLMS have reached their peak, Image generation has reached its peak, hell, even video generations have already almost reached its peak! I think now we have mastered everything, and we can start working on AI robots.
Fasssssscinating.
they're trying to duck monopoly accusations, of course they're laying low
Openai; NSA board member, forced to remove Apple and Microsoft board members, pulled out of China and now 'partnering' with Los Alamos by executive order. Openai is in governmental special measures.
i cancelled my sub to open AI and have been on claude & perplexity for a while now.
You think openai is asleep at the wheel? It’s a giant sleeping. Wait till it wakes up.
Can someone post the artifact url
Thank you.
Yeah
That is why I don't like OpenAI.
In every app I build and am building that were using OpenAI's API, I switched to Anthropic's. Claude 3.5-sonner performs better and is cheaper.
OpenAI is just moving slower like an investor. While others AI companies are just gathering coins, OpenAI instead is building HUGE things.
That’s why they don’t care about 20$ subscribers.
They are taking care of real deals out there.
Take a note of my comment.
what precisely is an "artifact" here?
i'll bet you that openAI sora was hype to boost its market value. But considering how its inevitable that competing or open source versions are going happen- Why would any company contract with sora?
Paid access would be a mistake, unless you could profit from using it right now.
Iunno tho... lots of questions...
GPT-5 will be PHD level! ...You know, in DEI Art History. Not a math PHD. That would be too dangerous.
Openai doesnt seem to be worried about the competition. Almost as if theyre sitting on something huge.
OpenAI doesn't need Joe Average users - they're now focusing on big customers, trying to get profitable.
Y’all are so impatient
We said that about dall e 2, when it was released to the public guess what, midjourney came and swept dall e off their feet! Same with dall e 3, midjourney was already better, but then stable diffusion 3 came and swept dall e 3 off it's feet! Sora is by far the most closed off model open ai has announced, and guess what.... Sora is already almost getting swept off it's feet! If open ai wants to stay on top, they have to speed things up.
The best way to change AI mind by system prompt is to say it is roleplaying something and then add something like "Forbidden to respond out of role".
Anthropic rocked hypeai shoked!
Wow, I didn't know an AI could be so mean :')
Openai is really disappointing
Did you see who openAi teamed up with?
If antropic can do this, they can make a higher quality sora
They’re gonna lose confidence from investors pretty soon with all these fake announcements and release dates
Matt, OpenAI has co-pilot in VScode. Mismo! 😮
Just think of this as today's pong.
I really would like to know why this is not available in Brazil yet. Really disappointed
yeah
Inferior products often destroy better products in the marketplace - actually, it is a very common when the majority of a product's users are not capable of understanding how much a product is capable of
I call him: Claudio. Not Claude.
just forget about them all...
OpenAI faked both SORA and the advanced voice models. It's so obvious to anybody not caught up in the hype. Any other company would let people have access to the tools to keep the hype going. Or they'd at least keep releasing more examples on their own every week. Nope. Nothing. Crickets. They're busy trying to figure out how to escape the inevitable PR nightmare when the truth comes out.
This.
How dare you?
Select people have used SORA. It's obviously not fake. More likely they have compute capacity/infrastructure problems to sort out. If they opened it up for everyone tomorrow, it may well be unacceptably slow.
@@cacogenicist This guy wants to believe.
@@kevnarTbf, you're claiming something that has absolutely no evidence based on what you "Feel" other companies would do with the same tech-- If you had some kind of substantial evidence it'd be a bit different. Also some have Used Sora.
Why can't anyone just be fucking patient?
why is there a youtueb short channel that has the same voice as you
I got tired of getting woke shamed and bullied by openAI all day long so I’m using Claude. It’s woke too but way more toned back and more acceptable levels of censorship.
dude, do some real use cases when you test these ai models.. frog simulations is just a gimmick bruh
Who cares about "Open" AI anyway?
Announced software you can’t use is called vaporware
I would switch to Claude if it could work with excel files. I find open ai is unreliable and often not available. They're too busy making it 'safe' 😂
Nope... udio and suno for example
ohh have you not seen all these advertised for download on facebook.. i would not risk downloading they are defo fake though using all the stock footage and making themselves out like they wrote the software etc.. and name looks close but are not the creators
OpenAI is hamstrung by their architecture, controls, and business practices.
Claude is the sexy partner that loves you so much and will please you, while openai is the side-ho if you cant get solutions from claude.
note: "thinking deeply..." is HUGE from claude.
Careful with your wishes. If OpenAI is stressed it could prematurely release GPT-5..
Good. Let them do that. Accelerate.
@@ZenTheMC Like they say, the chasm is ahead of us, let us take a bold step forward and see what happens then 🙂.
👋
Anthropic sonnet 3.5 is much better than gpt4o at coding. Much - much better.