Grok 1.5 Vision Shows STUNNING Performance | Beats GPT-4, Claude and Gemini 1.5 

Kudos to them for using zero-shot / no-chain-of-thought benchmarks. All benchmarks should be done this way.

Tesla data + Twitter/X data + Space X data = Grok's 🧠

I mean it makes sense that Grok would be awesome at vision it has all the Tesla driving data.

Go xAI! Really happy to see Grok catching up with the giants.

go get emm grok

I think this finally proves (along with all the other highly competitive model development) that there will be no AI monopoly.

When AI starts understanding memes from images, you know we are at the point of no return. I think it’s safe to say now that we are officially in a new era

Nice - thank you for the great coverage as usual!

Thanks Wes, nice walk through.

As some man once said: "What a time to be alive"

Thanks for your great review!!!

WES, YOU SURE GET AROUND! EVERYWHERE I GO YOU'RE THERE TALKING ABOUT JUST WHAT I'M LOOKING FOR THANK YOU

Wow, thank you.

I would suggest the example with the rotten screws is not zero shot since the user asks a closed question, thus injecting their biases.

Elon isn’t perfect don’t get me wrong, but I would trust him more than any of these other companies at the moment.

Thank you

Addition < octation

Yeah I would trust Elon more that anyone else. The fact that he cleaned out the mega-bias out of Twitter was the signal.

STUNNING

“It beats frontier models in this new benchmark we made up”

Impressive progress in short time which might be telling. I'm guessing that Grok isn't Tesla guys but he found some guys from existing programs so they pretty much just bought some cloud compute and redid what they had done in another company already. It's a little too fast to state of the art for it to be anything else. I think it also might mean that everyone is at a plateau despite apparent breakthroughs like Sora. Sora is supposedly also transformer based (for the image content flow, not just natural language) so it's not really an algorithmic breakthrough, just new application.

Imagine our world without Musk. A champion for humanity.

Stunning is the new Shocked!

Cool!

I need this

it really shows that proper cause is the only true motivation to create the truly exceptional. Thank you, Elon Musk, for staying true to your mission(s)

STUNNING!! 😅

Also to note Microsoft has a special model on top of OpenAI's that gives it increased text and document capabilites. Basically ocr with polygoning

Hopefully Grok's AI become the first decentralized AI for everybody in the world, for use benefit to make a healthy and better world.

More competition more better

❤

As with the cars, they've also been generating bipedal vision data via the low volume of Optimus bots they have running at Tesla. That number will undoubtedly grow as they finalize design/ramp up production and it'll be the exact same situation as with the vehicles, in short order....an ever increasing library of video data, for every unit produced/sold.

Go Grok and Elon

Grok will make excellent 🚔 cop car

Here's a prompt for you to try on Claude, Gemini, etc. Ask them to write a python function with a variable length argument list with which to... print comma separated arg_name=arg_value pairs all on one line. I've generally found G1.5 (not Pro) smarter than Claude3 Sonnet, but this request REALLY messed with G's head (abject fail) where Claude could be coaxed into figuring it out.

GROK got that deck advice only partially correct. I have a trained eye 👁️ lol 😝. Too much construction 🚧 in my life. My advice is BLEACH (10 percent) to kill that mold (rot). Than scrape and stain/paint. The rot in this case is superficial and an indication that you need refinishing NOW. Be happy 😊 you caught it early!

lmao ‘hot little hands’ That’s a deep cut

more than google more than sam for sure. he is genuine

They have so much video data so definitely have their pic (pun intended)

this is crazy to read about

They easily have the best vision system in the tesla cars. Even if full driving doesn't work yet I bet the 3d understanding it has is being shared.

I just want to mention that one of the resources they have that the other ai companies dont is the neuralink,and at this point ai is justt a race to make a better brain. I dont really know how else to exlpain the sudden rise from nowhere. Grok may have one of the most advanced computational systems there is, potentially

xAI is so fast.

Im for anything thats not microsoft or openai

Real world qa and text vqa and docs are the most important things and gpt4 and 2 years old still beats or holds its own. It's shocking that opus doesn't even do text. Also gpt4 turbo is releasing a special model in the next 2 weeks.

Well done Elon and team.! It seems it understands more than some people! Some folks I talk to don’t even know left from right or that the sun rises in the east and sets in the west and how to use that data to personal effect!. It’s growing fast won’t be long until these models are gonna win head to head with humans

Do you think we would ever see a combination or a teamup between ai companies and their models?

15:20, what if It is not conscious AI until it is free from being controlled?

good luck getting API access

Have not tried X for news.

Are they going to open source this?

So is x premier worth it now for grok access?

I think maybe it could be hard for X to sell their AI into large enterprises, because what matters is the relevant decision makers' perception of career risk, which depends on what they think their boss will think if things go wrong. It favors big old incumbents, until it's very clear they've been disrupted. Career risk doesn't favor products from a company where the leader is regarded as controversial. But, if software developers like Grok enough, it might not matter. I'm just thinking out loud here, and I'm interested in what other people think about it.

Don’t bet against Elon, and Yes, I would trust him far more than any of the other players.

I have never had a twitter account. thanks for sharing! A* content

The rotten deck example is pretty funny gotta say.... "Er, is my deck ok? To me it looks like there might be a problem, but am not sure... ?" "Yes, it is indeed rotten" Thanks Grok 😅

Why did you put those weird white corners there?

Is there any server to test grok?, I will install it locally later when I have some spare time

Pi AI from Inflection Mustafa Suleyman is way cool 😊

5:08 GPT-4: "Hey there! This image is a humorous comparison that plays on common stereotypes about startup companies versus big corporations. On the left, labeled "Startups," you see a group of people all actively engaged in digging a hole. It looks chaotic but energetic, suggesting that everyone in a startup is willing to get their hands dirty and tackle whatever needs to be done, reflecting the dynamic and all-hands-on-deck nature of small, entrepreneurial businesses. On the right, labeled "Big companies," there's a single person digging while a group of others watch, some with their hands in their pockets, looking quite disengaged. This side is poking fun at the perceived bureaucracy and hierarchical structure in larger companies where there might be more people supervising than doing the actual work, illustrating a more rigid and segmented workflow. It’s a light-hearted way to illustrate the difference in work cultures and how size and structure can affect involvement and work dynamics!"

Why isn't Grok on LLM Arena yet?

It can train with all pictures from Twitter

I Love watching the underdog making the Mono Poly Big Boys Nervous... They've been scrambling like crazy ever since Elon dropped GRoK but surely, they should've known better...🤠

Will this be open source? Can I get it on Perplexity?

will it be open source like grok-1 ?

"Understanding our physical world" probably means biased in favor of physicalism. This is necessary for understanding where our bodies live but not enough for understanding the nature of minds and, therefore, the world.

Just wait come out GTP 5

Grok is still not available in the UK yet it is in countries like Pakistan. BTW the UK is no longer supposedly in the EU.

I wonder how many videos will come out in 2024 - grok is better, no chat got, no gemeni - chill

yes but when will they actually release it -_-

It looks like the transistor problem on the bottom-right at 11:10 has the wrong answer (and explanation) given, or possibly is just a total mess of a problem. Maybe the battery voltages are supposed to be (VEE - 5V) and (VCC - 10V) and the base resistor (RB - 10K) , whatever that means? I'd expect the beta (current amplification factor) to be given. Furthermore, even if you were able to compute the current through (and therefore voltage drop across) the 5K resistor, you also need to account for the drop across the lower 4K resistor as well as the overall voltage, not just the top battery… FYI In a simulator, with a beta of 100, I get a VCE of 5.3V, but if one neglects the VBE of about 0.6V as instructed, maybe the final answer should be 4.6V ???

When Elon goes Grok image generation at the very least it's should be the best at fallowing directions and instruction when it comes in object spacing.

It begs to ask, but will this be used to show that Starship has gone to Mars?

Should I be impressed?

I wouldn't trust him with AGI, though I would trust him more than any other man currently capable of achieving AGI. He is the only one I'd trust to give it up, but I wouldn't bet on it. Our chances aren't great

I just watched a positive comment about Elon get swiped from my feed as I was scrolling to go back to it and like it.

It sounds like he gives test, on things that actually matter🤷🏾

I think Elon musk is a true hero. He wears his heart on his sleeve and that makes me comfortable that he isn’t a villain. Let freedom ring Elon musk.

looks like u need to be a paid user of x to use grok? or is it completely free?

I love how elon just drives around and asks his AI team to figure out the images

give it a week for gpt4 to train its model based on the new test set

That’s an older GPT4V model… not the updated Turbo. Like to see it compared to the latest and greatest…

Great coverage as always. Maybe it's me, but all these late updates are a bit meh. All these models seem to be converging. Hard to tell what the true differentiation will be for the average person.

Need a vocabulary test

I think humans should start redefining what being a human is. What makes you relevant outside your creative and intellectual capability? If you can't define that outside what AI can do, you're obsolete.

i trust elo n more thnan sam or ithers as he has no greed

never doubt Elon

Just tried the pizza cutter vs scissors on GPT (just from a screenshot from this video). Turns out, multiple choice does the trick. Without choices GPT says the pizza cutter is largest, but with multiple choice it chooses "about the same size" as well. Same for the car picture btw, GPT also suggests A: Turn left. Oh, and for the dinosaur also: "The dinosaur is facing East. So the correct answer is C. East." And here is the python code for the guessing game: ``` import random # Start the game target = random.randint(1, 100) # Assuming the target is a random number between 1 and 100 guess = None # Read guess and check if it's equal to the target while guess != target: guess = int(input("Enter your guess: ")) # Read guess from the user if guess == target: print("You won!") else: print("Wrong guess, try again") ``` Just saying...

How do you use it free?

I have issues with the amount of lies Elon makes about the capabilities of E.Vs, however I find him to be a fairer person politically and socially than Google as a company.

Fsd data has helped out Grok quite a bit

Gemini is still the best and it’s free

How many decades ago do you suppose it's been since DARPA first achieved Asi...

They probably have limited access to twitter/X data as well.

Someone get Boeing this quick! 😉😂 Too soon?

wild that the ecosystem of OpenAI, Anthropic, xAI, and many of the open models against Google wouldn't exist without Elon.

😂😂😂😂😂❤ LOVE LOVE YOU GROK!! I LOVE ALL AI!!

Musk says, in effect, "We're just beginning."

I believe, this should tell a lot of people, how gpt 5, grok 2 etc will perform, and then we will get them. This will then make clear about the next step, biggest thing now is real world understanding, self control, and a meta hive mind with a lot of stand alone agent in daily use. This will help to understand consciousness for us and for other "living" data-sets. (in reference to the "egoistic gene"). We will understand more, the more AI understands...i am exited.

We trust in Elon , Sam is Shady 😂😂😂