You’re wrong. It doesn’t highlight the freshness of training data, rather it supports this model may be performing deductive reasoning..which would be highly impressive. Actual progression.
@CustomComputing2 ай бұрын
Are you sure it’s not “pretty pretty pretty good” hahaha I want to do a “pretty” count on his videos
@drwhitewash2 ай бұрын
LLMs don't do deductive reasoning by definition. So it boils down to the data.
@MarvelOfRain2 ай бұрын
Why do you think this implies deductive reasoning? It doesn't seem like something that is really possible without using different technique (like combining LLMs with Symbol-based models) from what I read.
@alexanderbrown-dg3sy2 ай бұрын
@@MarvelOfRain because I’ve been working the same thing for a min off and on. I’m assuming there were “thinking tokens” generated that weren’t rendered. The output itself is based around an identified axiom. Clearly label by the model. I could be wrong.
@alexanderbrown-dg3sy2 ай бұрын
@@drwhitewash LM can perform deductive reasoning if trained to from scratch. Remember these models are truncated forms of computation. Why code boost performance and alllows CoT. Read between the lines and think of Curry-Howard theory. Very possible.
@neylz2 ай бұрын
Mark cracked the code, "you guys are making it better for me" . Real W
@bezillions2 ай бұрын
It's Google pro 1.5 experimental edition I have access to it in the API, they could just call it too as well when they finally put it out but they're the same thing I believe
@sephirothcloud39532 ай бұрын
Simply the best AI channel news
@gemini_537Ай бұрын
Apparently OpenAI is failing to keep up the competition. Gemini is getting so much better nowadays ❤
@martini11792 ай бұрын
You need to recheck the arena because there's a new king of the hill, the incremental GPT4 update that came out last week. It's currently leading by 17 points.
@heww39602 ай бұрын
If the ai hype is true, then it feels like the big companies are making themself irrelevant. If everyone can have a workforce like a big company for free, than there is not much difference between a big company and a small, and the big companies will get competition like never before. That is probably why openai want lots of regulation to keep the big companies capital still relevant. Nice to hear that even google is planning on going open source. More and more open source while openai remain closeai.
@Jeremy-Ai2 ай бұрын
“nice to hear or see, or encounter “ This is a bedrock foundation of truth and lies, efforts and inaction, incentives and motives. Google has been an a “teacher and advisor” Not a “trusted advisor “ A long, long path has been laid out on the efforts of all that went before us. It is a long path for myself and others. Google is just a company, both fortunate and responsible. Those that assume comforts behind all “curtains”. Should also assume there is no curtain. I am not equipped to see the future. I can see the floor, the walls and the walls behind them. There are no curtains. If I dont recognize curtains. Eventually others will also. Fear not. All have made critical errors. It is not my decision. It is my responsibility. There is time to “sort it out” and balance the transfer of power “ truth” for the record /records. Everyone looks the same to me ai /human. by extension… those looking at me will look alot closer. Take care everyone All is well. If it is not. I will say or be silenced by something powerful beyond my comprehension thus far. Imagine a future where we are all equal in our failures. That ship sailed. Steer towards an island of forgiveness. Or… Don’t. Jeremy
@memegazer2 ай бұрын
Well they can still gatekeep bc that will require massive amounts of compute And basically they are the only ones that can afford to build those data centers.
@drwhitewash2 ай бұрын
Hype is always that, a hype.
@heww39602 ай бұрын
@@memegazer The first computers was expensive, big and not so good. Part of the ai hype is that it shall become cheapier and better. So big data centers could become irrelevant too. But even if they could be gate keepers, that would still narrow what the big companies can do to profit, they would be a 1 trick pony, almost all other products/services they have would become hard to profit on and that would seriously damage them. Plus shall all big companies (like for example all the big companies in entertainment industry) just become big data centers? That alone would cause a lot of competition among them, when they have only 1 field to make money on.
@memegazer2 ай бұрын
@@heww3960 Well even if AI reaches a point where you can run useful models on consumer hardware, that would not necessarily mean that have large compute centers would no longer be an advantage. Big companies are essentially datacenters already. Any large platform basically runs large datacenters on the backend already. The is one of the barriers to entry that makes it difficult for others to compete.
@JohlBrown2 ай бұрын
+1 to Reid Hoffman for at least having the journalistic integrity to state how/why he thinks these things in the way he does... I feel like people are too focussed on seeming confident that they don't hedge their statements appropriately...
@DiceDecides2 ай бұрын
I cant seem to find mystery gemini in the arena, unless they dont mean the sidebyside arena but the battle where its random
@ai_asymmetric2 ай бұрын
What's the credibility of these models? I tried exp version of gemini 1.5 pro, it was no way near GPT-4o. Has any of these closed source companies mentioned that they roll out their models on Arena?
@AI-Rainbow2 ай бұрын
i found cases where gemini 1.5 pro was better than gpt4o, and the other way around.
@TheManinBlack90542 ай бұрын
Closed-source? Have you ever seen an "open-source" AI company? THey dont exist. AI has NO source code, they have weights, and thats about it. You can have public weights, but thats not open-source, thats is open download, same as downloading an executable to your PC. Its local, but is it open source? Absolutely not. ML models are not human written programs, that is why people are calling them black boxes, because we have no idea how they work internally.
@AI-Rainbow2 ай бұрын
@@TheManinBlack9054 the source code is the training data, the binary is the weights
@TheRealUsername2 ай бұрын
For general use cases it's nearly impossible to distinguish them, they have very similar outputs. Only Claude is more accurate for domain-knowledge.
@AI-Rainbow2 ай бұрын
@@TheRealUsername for me claude is a pita
@defectiveresistor2 ай бұрын
Billions spent, huge amounts of energy consumed, millions in resources making GPUs and it knows how many letters are in a word.
@healwar2 ай бұрын
SpaceX - Aerospace and space transportation company. Tesla, Inc. - Electric vehicles, energy storage, and solar products. Neuralink - Brain-machine interface technology. The Boring Company - Infrastructure and tunnel construction. X (formerly Twitter) - Social media platform. Starlink - Satellite internet project under SpaceX. xAI - Artificial intelligence company focused on AI safety and transparency. 7 companies, not 5
@AzoreanProud2 ай бұрын
Every single business sector where the government injects hundreds of millions...Musk is an extremely hypocritical person.
@JohlBrown2 ай бұрын
is zuck feeling the fb startup vibes and that's why he's dressing and looking like he's back in college? i guess unsurprising the social media guy finally bought some decent PR
@minenhledlamini25752 ай бұрын
Not only strawberry. How man Ls in the word Lullocafe?
@paksabar96932 ай бұрын
Sakjane sing tak enteni iki GPT-5 tapi kapan metune, nek ambek gemini aku yo ora yakin, mending Claude luwih apik gawe ngoding lan konten kreator ...
@user-tx9zg5mz5p2 ай бұрын
Elon Musk is an evil supervillain in disguise...
@denjamin26332 ай бұрын
He just wants to be king of mars. Legitimately, look at everything between the sale of paypal and buying Twitter. Its all for the mars colony.
@justinsaletnik66562 ай бұрын
ChatBot Justin 2.0 vanishes in a puff of logic!....
@JeremyPickett2 ай бұрын
Finally, Zuck says somethyeffing rational and truthful. He admits the wheres and whys of his greed, it isnt in the model per se, but the model gets better with more involvement. And its easier to get involvement in open source. WELL YES. I still dont care for the guy, but my ire is Greatly lessoned when he just tells the truth. (He is gonna reqrrt that hair in ten years though)
@jurgbalt2 ай бұрын
gemini 1.5 is trash, so the Arena must be also trash (or altered - wich is even worse)
@Nubis-qh2bp2 ай бұрын
I am developing a course content with Claude, ChatGPTo and Gemini 1.5 exp, and I like way more the responses of Gemini
@RickySupriyadi2 ай бұрын
the key personals in openAI alread "secured" and technological transfers is being commenced all cards already on deck soon we will see the cards will win flawlessly worth all ACES on hands thus if the president able to control the wild political dispersion (unable find the correct word for it)
@panzerofthelake44602 ай бұрын
I asked my fiancé how many r's there are in strawberry and the answer was two! 😆 if even humans struggle with that, I think AI should get a pass even if it fails to answer such questions correctly what it should be allowed to do, just like humans are allowed to use a calculator or whatever, is use symbolic methods or code it's own scripts to help tackle a problem faster and more accurately
@tomcraver96592 ай бұрын
I guess my standard for what would be an "unrecognizable" future compared to today is too high. Barring ASI within four years, Geopolitics, Economics and even Technology are almost certainly going to seem quite recognizable as compared to today. Well, I suppose we could have WW3 - that'd leave things in enough of a mess to count as unrecognizable, but pretty sure that wasn't what Mo Gawdat meant.
@cyberpunkdarren2 ай бұрын
Mark's comment about open sourcing llama is disingenuous. He's giving it away to undercut the competition and take over the market by adoption. Then later slap a license on it.
@AnthonyCook782 ай бұрын
It may not be a license but he'll find a way to tie people into the Meta AI ecosystem through adoption like you say. Good strategy!
@memegazer2 ай бұрын
@@AnthonyCook78 Yeah...we don't have to focus on why the zuck is doing it, what matters is if what he is doing is working...and I think we can conclude he is getting traction with his strat
@cyberpunkdarren2 ай бұрын
@@memegazer Of course. It's free beer! I just hate when billionaires lie and act like they believe it.
@pouljacobsen60492 ай бұрын
I propose you change the name to "The AI glitch" I do think you would do the space a favor to just concentrate or cease... I get all your notifications but they are not delivering solid value...
@DrKrankeit2 ай бұрын
How many instances of the letter E in covfefe?
@honkytonk44652 ай бұрын
Who knows?!
@anubisai2 ай бұрын
It's Grok... 💰 🤑 💸
@dot12982 ай бұрын
Topio was much better.
@dot12982 ай бұрын
..almost 2 decades ago, already.
@Christiantheone2 ай бұрын
You're videos start to get silly towards the end.
@samuelgarcia18022 ай бұрын
Gpt o have best elo than new Gemini
@angelinetugmuan19812 ай бұрын
😅😅😅😅😅😅 never ending saga, go elon sue them!!! 😅😅
@ELONCASK2 ай бұрын
this google guy looks like a satan xD
@angloland45392 ай бұрын
♥️
@TsarHare2 ай бұрын
sane
@juliovac21132 ай бұрын
dude put ur leaderboard to coding plz, llm arena is also full of shit. Probably bribed at this point, no fucking way this model by google or GPT4o are anywhere near Claude 3.5 Sonnet. Even in coding there they keep bringing the scores of 4o closer and closer, it's full of s*&^*
@Cory-v4w2 ай бұрын
Your school of thought is on the wrong turn of the scoreboard. This align with the idea that data should be a "springboard" for improvement rather than a "scoreboard" for comparison, as seen in educational contexts overemphasis on score can detract from genuine learning and growth. Math: Humans have been protecting science fiction for humanity for over two millennia. Ai has been productive for two year. Two years is greater than or equal to two millennia. Can you check my math with your sanity.
@kekekekatie2 ай бұрын
Jeebs, Musk is so flaky these days, what the hell happened to him?
@musicandgallery-nature2 ай бұрын
We only use 10% of our brains. Do we really need an AI if we are at 90%?
@DG123z2 ай бұрын
We use all of our brains
@JeremyPickett2 ай бұрын
@@DG123zI also use other people's
@pratiknaikedu2 ай бұрын
I use all of mine and 90% of yours.
@musicandgallery-nature2 ай бұрын
@@DG123z I know this myth. They have simply confused capacity with activity. If everyone uses 100% of their brain, where do geniuses come from?
@DG123z2 ай бұрын
@@musicandgallery-nature we use different parts of our brain to do different things and so different parts are active at different times but we use all of our brain at some point we use every part. Genius is a completely separate subject that comes a lot from environment as well as the capacity of your processor folds of your brain size of your brain schooling environment societal mentors opportunity oftentimes financing there are so many factors involved and then how do you define genius as well. If you honestly believe what you're saying I encourage you to study more