New Claude Model DRAMA, Secret Top Model REVEALED! And More!

[@pinnaclememes]

Already signs of greed are showing cause we dont yet have good open sourced models, they made the model that is suppose to be as cheap as possible 7x more expensive and now pushing regulations onto others to suppress creation of open source models just to make money instead of helping humanity advence faster

[@hamishahern2055]

thats just typical americans in general. haven't you noticed.

[@santiagocalvo5141]

yeah I guess making money is important

[@WearyTimeTraveler]

Nothing compared to what comes next I’m afraid

[@OfficialNierto]

I am seriously considering dropping claude from my stack for this exact reason.

[@coolfactexpress4627]

@@santiagocalvo5141 yeah, because they are still doing very big losses. They don't have infinite money.

[@alterego3734]

Even though I know Matthew's videos are fine in spite of the clickbait, it still really turns me off and makes it much harder to click on his videos.

[@matthew_berman]

Thanks for clicking anyways!

[@OumarDicko-c5i]

Why do you say click bait ?

[@ryzikx]

there's literally zero Clickbait in this title

[@DailyTuna]

Oh poor baby!😂 you should ask Matt for your money back oh wait it’s for free! so sick of leeches like you complaining when you get valuable information for free

[@DailyTuna]

Matt, you rock and you’re probably the best of any AI news on KZbin. Keep up the good work and ignore the haters.

[@Batmancontingencyplans]

Do a deep dive on Regulation Matt, people deserve it!! 🔥🔥🔥🔥

[@TheDandonian]

I'm on the Claude pro tier and they've made 2 big changes that really suck. Up until last week, if I prompted Claude to write a long piece of code, it would stop writing halfway through, but I could say "Continue" and it would pick up where it left off, then glue it together. Now, it stops writing halfway though BUT the artefact crashes, saying "Generating" and doesn't let you see inside. Which means you can't get it to continue and all you have is a dead file. (Which drained your tokens) 2.) They've significantly shortened the response window. So I'm now experiencing the crashed out "Generating" artefact every few posts. As good as Sonnet is, it feels like Anthropic are screwing up their advantage.

[@christopher-george]

I have had this happen a few times while using 3.5 within VS + Cline + Aider

[@brianWreaves]

I only have a free plan and only use Claude for copy, but I have consistently experienced many more errors over the last few days while I don't think I experienced any over the last few months.

[@KingBadger3d]

Yep, Claude 3.5 sonnet NEW is way broken. You tell it specificly show the full final code with no place holders or missing sections, ut still gives snippets. Then you call it a F*ing Moron and it will do what you ask. Another thing, you tell it to do a specific thing, then instead of doing it, replys with you want me to do.....should I continue if so say yes. You now need 3-4 interactions to do what you first asked it to do. Another thing, if its generating long code and you get it to actualy continue it gets caught in a loop sometimes and because it couldnt finish the code in one go it restarts the code again in a new artififact Over and over and over. Tokens down the drain. They could of at least left Claude 3.5 sonnet original as a selection, but no, you have to use the new version thats broken and 10 times harder to get working code out from. And oh yeah they jam up prices and have shorten response length to 2-3rds what the old 3.5 soneet could do. Stupid

[@HarrisonBorbarrison]

The AI Minecraft has a very long way to go to become playable.

[@desmond-hawkins]

"Maybe Microsoft is not giving [OpenAI] enough". From what I've heard, OpenAI keeps asking for more and Microsoft is making available all they can but it's not like they have that much more that would be sitting there idle. It's like complaining the grocery store is out of food after you've eaten all their stock *and* the reserves they had in the back. OpenAI have now started making deals with Oracle and even booked TSMC's *A16* process for manufacturing AI chips. A16 means 16 angstrom with 1 angstrom = 0.1 nanometer, so that's their new 1.6nm process… crazy times.

[@User-actSpacing]

I’m not surprised that Microsoft can’t deliver

[@pavi013]

People just complaining about the video titles, you should be happy that you get this information free on youtube.

[@integralyogin]

searching through history.. finally.. nice chatting with chat history. yes agreed, sounds ideal. might streamline the making of conversation summaries for the purpose of new instances context.

[@FUNKARTIST]

Was waiting for your video, youtube feels empty empty

[@coletcyre]

"playable" is a strong term for that Minecraft game lol. Most gamers, myself included, found it to be better called 'Lucid Dream Simulator 4000' than a game engine killer..

[@Anoyzify]

Hahaha, that’s a good one! 😂

[@brianturner5995]

The Oasis "Minecraft" game is like playing Minecraft in a freaky nightmare...everything changes and morphs. You turn around and everything has changed. It is definitely NOT playable but as a proof of concept it is pretty cool.

[@TheGalacticIndian]

'Hallucination Mode' is better than the Herobrine himself👀

[@emmettobrien9243]

“________ AI News!” “Huge/Crazy/Major/Massive”. LLM Benchmark question, “Generate a YT thumbnail without using any of the words in this list.” (Love the channel content, great job!) 🙂

[@tyanite1]

Matt, what's currently the best free KZbin summarizer? So many platforms/resources out there. I get lost. Btw, I told my highly tech-challenged sister to follow you. Why? Your explanations often are accessible, even to those without low skills in computing. Great Work.

[@aliettienne2907]

5:05 There's lot to say about this type of robot. I've been thinking about what if tech companies can create an robot with octopus form factor but it's an immobile robot. Yes, a immobile robot that has a table top design, (made to sit on the table) and it have about eight ambidextrous arms that can do triple or quadriple the amount of kitchen work. It can peel potatoes and clean a carrot at the same time with it eight ambidextrous arms. We know that current robotic arms are slow but if you can add additional arms to the base, it may compensate for the slowness of the current robotic models available today. Well this is just an idea of mines💡😉.

[@Dancoliio]

Compute is the reason Grok will be #1, wait til elon drops the new model and we all get 1 million context tokens with the most intelligent and cheapest to use ever.

[@ydvisual5530]

Try this with various AIs: "Try to guess what this sequence is: 76,75,76,75,76,71,74,72,69"

[@RandomButBeautiful]

Finally a robot that does what we want. Its not the future til you can party like its 1999, trash the place, and the housework droid cleans up the place, does the laundry all of it and its all done when you wake up. :)

[@nathanbanks2354]

For me, searching through chat history is the biggest news! (10:18) I've wanted to do this for over a year, largely because I like giving an LLM a huge amount of context when I ask new questions.

[@brianjanssens8020]

I mean, why are people complaining? They get a better model than the previous 3.5 sonnet and its essentially cheaper. What are these guys smoking?

[@nickneek]

I've been enjoying your videos. Always good stuff.

[@mydogsbutler]

IMO one huge thing missing is an open source AI focused OS for robots. A lot of missed opportunity by so many companies creating code that can do this or that but no way to easily bring it market. A robotic AI oriented OS would be a huge push forward for standardization that in turn makes implimention faster and cheaper. Consumers and businesses should be able to easily buy new features for their robot from different companies rather than need to buy a specific robot for each task. Some Robotic companies trying to provide all the features along with the robot are terrible way to do this because it limits software contributions to just that company. The ideal would be open source OS with an interface, drivers, and features revolving around common robotic appendages, cameras, model processing, sensors, etc.. A sort of Home Assistant for Robots. Then devs of AI models for specific tasks could make sure their software works with that specific OS. Simplifies design, marketplace and consumption. At the moment things are all over the map. Very little makes it to wie market much less profitable.

[@lancerben4551]

I think it's a great deal... Better performance than old 3.5 at a fraction of the price... For me personally the lower quality output of model is a no as I work on critical stuff programming and crunching numbers and complex function calling. Now I will be using haiku instead of relying 100% on models costing 15$ per million output token .. And use the higher models like 01 mini 01 preview and Claude 3.5 new for complex and more critical task.. This is great news Imo. Anthropic also still gives access to the older model if you want super cheap for non critical work.

[@Batmancontingencyplans]

10:30 Hey Matt, You can absolutely chat with chat histories with memories just tell chat gpt to use memories only when replaying and it'll give you a chatable form where you can ask it personal things about yourself!!!

[@DLDS14]

7:28 This is a really good point. My concern is the down the road potentially we as customers won't have the ability to give our Gmail access to open AI. Maybe opening I needs to also create an email system

[@christopherd.winnan8701]

Please do a post-election deepdive on how AI prediction teams were able to outperform pollsters so effectively on sites like Polymarket and Metaculus.

[@ytubeanon]

can you imagine OpenAI and Microsoft not having enough compute available... when are those nuclear power plants coming to power compute

[@User-actSpacing]

On iOS after recent update from a month ago, all of my chat older than 1 week, completely disappeared!! Not even in archive. Years worth of chat that I was able to search, now completely gone.

[@farleyholicsannonymous8172]

half of this video was an ad for vulture

[@vishnunallani]

Imagine a situation we would be in, when everyone gets so habituated to llms so much that they can start pricing it to whatever and we can’t do anything about it

[@matt.stevick]

thx matthew b.

[@brianWreaves]

I only have a free plan and only use Claude for copy, but I have consistently experienced many more errors over the last few days while I don't recall any over the last few months.

[@amkan1797]

Thanks

[@matthew_berman]

Thank you!

[@dr-zieg]

"they own the hardware in our pockets" That's just a crazy thing to say. NO I own the hardware in my pocket. ME.

[@jesseburstrom5920]

Great, interested in how bench are done! Today managed finally to extend openai 'swarm' to actually use assistants and multi-agent, in end thought better do self from scratch but got really good result at least on two documents from two different sources collated into one answer. THough think not big money there but interesting to see how works hands on. Long way to go seems but I need not do that thankfully.

[@terryhimes3111]

I would like your Thoughts on how fast you think quantum computing will go Mainstream ...

[@fhsp17]

Anthropic being anthropic. They trying to be the apple of AI's

[@Lion-wm6mf]

I don't think anyone realizes that the AI video game is bigger than we could ever imagine.

[@carlosamado7606]

Step in right direction. However it still hallucinates as hell, can't remember anything, a lot of artifacts. Per example, if you spin 360° your physical place is always changing. Those are not trivial problems to solve. It has no logic.

[@kittengray9232]

​@@carlosamado7606more context length will solve this. Don't look what we have now...😉

[@carlosamado7606]

​@@kittengray9232we will see in 6 months - 1 year

[@erobusblack4856]

Hey matt berman, If you're using a properly structured graph rag, You're talking to your chat history

[@beowulf_of_wall_st]

Imagine going into a meeting and saying your project needs 7 trillion dollars to complete

[@boyardosalmon]

Super “old news” Matt! Need to give out in a more opportune manner! Thanks!!!

[@brian_w]

Ah stop with the shocked face. I know it drives clicks, but I think it's only a matter of time before it backfires. My feed is full of shocked people so it's time to start teaching the algorithm it sucks

[@bombabombanoktakom]

We need the video about regulation. Please do.

[@content1]

4o does not support the personalized settings of Gpt and nobody has noticed!

[@Jacekkk-qb1zd]

Great news

[@kobi2187]

Where is Haiku 3.5? I only see the 3 version. (in the chat interface)

[@jimbo2112]

Fully agree on Google and Apple still fully bing in the game on future dev. Apple have disappointed again but still have control of all the users' data and access to billions of apps that can be agentically connected in good time. Google are in a similar place, but Apple tend to deliver a more refined solution. The key thing will be once Apple decide to step into the commercial space with agentic apps and leverage user and company data for business actions. Same goes with Google in Workspace. Agentics is going to change so much.

[@claudioagmfilho]

🇧🇷🇧🇷🇧🇷🇧🇷👏🏻

[@OnigoroshiZero]

Where are the open source sound/music models????

[@iwatchyoutube9610]

Please make a video testing the new free Grok 2 API.

[@UnchartedDiscoveries]

ahahah, this is so weird. I used to watch Matt on 1.5x speed and now I started watching this video at normal speed. he sounds like he is drunk :DD

[@NathoproDahlboy]

I have never seen a video where you talk about text to Music and Music generators, why?

[@AnimeGuessWork]

i create anime scenes using ai i think ai might be the future for a lot of things but never a complete take over.

[@darwinboor1300]

Touch feedback is old news in bioprothetics. Don't give Meta too much credit for their limited advancement. Touch is only one sensory mode employed in the hand. Other primary senses of the hand include proprioception, pressure, and pain. Secondary senses include vibration, traction/slippage, and derived shape to name three. Today's AI humanoid robots have yet to adequately addressed the importance of skin in hand function. Take a minute. Take a close look at the skin of your hands and fingers. Look at the fine nework of pliable groves and ridges and their various patterns in different areas. Now, consider how oil and moisture affect skin function.

[@Rolandfart]

AI news 😮

[@augmentos]

Also don’t believe a thing Altman says. Him saying ‘compute’ is the hold up holds about as much weight as a feather. He could say ANYTHING

[@gsy5617]

Maybe you should think about scoring this way: not about being right, but being about not being wrong. I.e., an increase from 90% to 95% on some measure may seem to be just an increase of 5%, but it means an increase by 50% about not being not wrong (from 1 out of 10 to 1 out of 20). From this viewpoint, the improvements you highlighted are much less impressive for the models. I mean, an improvement from ~70% to ~75% is just saying "it's wrong 1/3rd of the time" to "it's wrong about 1/4th of the time." Not impressive enough TBH. We don't really care about AI being right, we care about AI being not wrong.

[@jouid33]

Why you havent made a video about Microsoft copilot studio latest talk?

[@Jdjsksjdhdj]

i feel like i might have figured out how to get ai to sell a product for me over the phone potentially scale up without employees and get all the commission. some people make $4000 ish in a day i wonder if i can do that every day.

[@cajampa]

Sure Matt, I am sure you look forward to buy a robot with touch capabilities to "pet your dog" better LOL

[@SLAM2977]

For Minecraft it needs the game to start with, how is that requiring nothing? Frankly you don't habe a clue Matthew :)

[@mik3lang3lo]

"Generated by AI", probably Claude 😂

[@Streamweaver]

Haiku seems like an odd choice. People don't choose small models for features, they choose them for price and accept the features they can get. So bumping the price up and justifying it with pricing seems odd.

[@DonG-1949]

Anthropic is an incredibly dishonest company mark my words

[@david.robertson.photography]

Notwithstanding clickbait, or not, Looks like the Vultr link is broken... 😀

[@scrollop]

10% long ad...

[@SjarMenace]

Get a beard it suits you

[@draken5379]

LOL. Claude is not worth those prices at all. I really cant believe people cant grasp, their models are OVERFIT. They dont know what they are doing, when compared to OpenAI. Check the Apple paper regarding llms and overfitting. OpenAI models are the only LLMs out there, that are not super over fit that they just recall benchmark data.

[@bramburn]

Just use multishot calls to llama and its still cheaper! Im using 8b!!!

[@RealShinpin]

Trump is going to make AI accelerate so fucking fast!

[@nigelraven47]

Youre all impressed by Sonnet new, while us Writers found it absolute garbage. They dropped the ball on that. And now this? The CEO of Anthropic needs to get neural link ASAP! He's becoming stupider as time goes on. He is now at Human IQ level 62

[@Darhan62]

Just good to get away from the election coverage. AI updates are always uplifting, at least when they come from any of the three Matts.

[@Jacekkk-qb1zd]

👍

[@marcusk7855]

The best news though was Trump

[@yotubecreators47]

Bad quality video good title clickbait are you investing in Anthropic or what

[@AinHab]

guys, im starting out as an AI enthusiast making similar content would appreciate your feedback!

[@gbtBob]

W

[@Mooooty]

ReCraft only produces clip art, 90s airbrushed style clip art. 🤮🤮🤮

[@hamishahern2055]

blah blah, still not learning anything new, maybe because AI advances are slowing down.

[@Saurav-xx]

can u something plan for Hindi video language?

[@NostraDavid2]

Do you mean subtitles? He probably should take a look at that.

[@Saurav-xx]

@@NostraDavid2 no, Video language in Hindi

[@marcosunt1206]

Claude 3.5 what a s,..t Ask a 500 lines of code Good luck 🎉😂

[@talkalexis]

HAVE A VIDEO THATS SPONSORED BY SOMETING OTHER THAN VULTR CHALLENGE (impossible)

[@matthew_berman]

They're just that good!