Gemini: ChatGPT-Like AI From Google DeepMind!

[@max-is-loud]

When it comes to academia, I am a layperson and honestly, kind of out of my depth. Dr Károly Zsolnai-Fehér is my go to for making sense of all of these amazing new breakthroughs. Indeed, what a time to be alive!!!

[@niismo.]

Exactly the same thing for me! It's way too complex and I don't even know what resources to use to keep myself up to date, but Dr. Károly Zsolnai-Fehér does the job for me and I'm just as excited about all these new papers and technologies, as he is. What a time to be alive!

[@greengoblin9567]

I tested them. Chat Gpt is way better. It is able to better extract information from its context length.

[@MaxenceFrenette]

same

[@lobabobloblaw]

Károly is doing a great job of staying above the tides, as wild as they seem to be getting. If I’m going to choose a ship to sail on, I’m going to make sure it’s made of two minute papers!

[@max-is-loud]

@@lobabobloblaw hold on to your papers!!

[@AMA14700]

I see that little AI is getting advanced so fast, more than expected "What a time to be alive"

[@emanu1674]

What a time to be alive!

[@FRareDom]

Hold on to your papers

[@igoromelchenko3482]

​@@FRareDomall your papers are ours. 🤖

[@-Timur1214]

I kinda want to say.. I was here Pog :D

[@daikucoffee5316]

What a time to be replaced!

[@igoromelchenko3482]

@@daikucoffee5316 😂😭

[@Megneous]

It's important to remember that Bard currently runs on Gemini PRO, which is roughly equivalent to GPT-3.5. Gemini Ultra won't be released until sometime next year. Gemini Ultra will likely be a paid subscription and will be Google's competitor to GPT-4.

[@gaweyn]

it could be competing with GPT-5 by then

[@hdtorresr]

Do not forget about GPT4 Turbo

[@justitgstuff5284]

Yeah, but it still has some nice benefits like being able to process media, which is a paid feature with ChatGPT. Integration with Google Workspaces is also surprisingly handy.

[@BeOnlyChaos]

Also bard has text only Gemini for now, according to the updates page.

[@justitgstuff5284]

@@BeOnlyChaos Kind of, it can't generate images or videos, but it can observe images you upload, as well as being able to create spreadsheets and documents. Kind of weird that it can't create stuff in google drawings though.

[@dasistdiewahrheit9585]

Next paper: AI needs your clothes, your boots, and your motorcycle.

[@knicklichtjedi]

I hope this will survive for a bit longer! Having more competition in the AI market is probably a good thing as ChatGPT (and GPT4) is quite dominant right now.

[@Tosslehoffe]

Unfortunately it's still not really competing with ChatGPT. Who knows how good the model itself is, but bard is still kinda terrible even with gemini pro

[@DavidBerglund]

I'm still waiting for access to GPT4 in ChatGPT so I've only tried GPT4 through Bing and GPG 3.5 on ChatGPT. But my conclusion from having tried Bard (before and after the upgrade) so far is that Bard is pretty great in some tasks like summarizing text and I'm able to have fairly consistent back and forth conversations with it.

[@DavidBerglund]

Granted, my anecdotal "evidence" can't be compared with more scientifically rigorous experiments like the benchmarks referenced by Google.

[@longjohnson3319]

What a time to be AI!

[@YoungMoneyFuture]

What a time to be Ailive😂

[@PuppetMasterdaath144]

I don't have feelings or consciousness. I'm a machine learning model created by OpenAI called GPT-3.5. I'm here to provide information and answer questions to the best of my ability. How can I assist you today?

[@JoeyJSawyers]

💀

[@igoromelchenko3482]

Stop the cultural AIsploitation.

[@luiztechai]

LOL

[@JohnDoe-rs4fl]

Based Petrucci. One of the greatest of all time.

[@para5maker]

Of course he listens to Petrucci, what a time to be alive!

[@Tesserex]

Yes! That solo album is amazing!

[@CreamusG]

I honestly can't wait for the time when it becomes good at math. That's so unbelievably important. Currently I threw a (relatively obscure) geometric problem at it (sum the outer edges of a number triangle), gave it a known solution (n(5n^2-3n+10)/6-2), and it said my answer was wrong and asked for a mathematical proof as to why their answer (2n(n+1)-2) is wrong. We've a ways to go, but the thought that we're even this close is insane

[@DreckbobBratpfanne]

There are some rumors about a new technology within these models that we probably gonna see in 2024 called Q* at OpenAI, it was a rumor at first but has been confirmed to be in development (and Deepmind has something similar it seems in the works) these could from what I heard help significantly with understanding maths and other areas. 👀

[@czerskip]

Not gonna happen any time soon. The complete lack of understanding the problems is what makes the "AIs" unable to do so.

[@programaths]

@@czerskip Even in Star Trek TNG, where they have very intelligent computers, the humans are needed to handle the subtleties. They also make a point of showing how stupid such a computer can be. One of the example is when a guy asks some water. The computer asks for the temperature and the guy is a bit furious and just say "Just give me water". In human to human interaction, such request is highly context sensitive. If you have some cold water, you would ask if the guest want tampered or cold water. If you have only tap water, but recognize the taste is not good, you would reply that you've tap water, but the taste is not the best. Here, the computer can create water at any temperature. It could be more human by knowing that tempered water is a safe choice. But it's not human, so it needs well defined parameters. It also always fails on idioms. Data has the same issues (and way samrter that the ship computer). It's quite an old show, but they were good at outlining AGI shortcomings.

[@ThePlacehole]

It's kind of weird, that people aren't numb to the constant hype. They keep claiming superhuman performance, and yet what we get feels more like CleverBot from 2008. Rare moments of brilliance, but mostly nonsense.

[@johnflux1]

@@czerskip "Not gonna happen any time soon" it's funny to hear people say such things with such confidence, when famously we already know that GPT-5 has solved the problem, and is able to do such math.

[@frankroquemore4946]

I think Google will compete, and maybe they’ll win, but I think a couple things rain on their parade that aren’t being considered… 1. OpenAI gave some beta users access to long term inter-conversational memory (like MemGPT), so that’s coming soon. 2. The footnotes of the Gemini tests basically admit to giving Gemini steroids while handicapping GPT-4 and then comparing those results (prompt engineering techniques for Gemini that weren’t given to GPT-4). An apples-to-apples comparison will likely reveal a less impressive accomplishment. 3. The version they’re showing off isn’t even available yet. And some of what they’ve shown off in their testing won’t even be available to users because it’s too computationally expensive. 4. By the time it IS available to users, I wouldn’t expect that OpenAI has sat back and done nothing in the meantime. Basically they’re a year behind OpenAI. The equivalent of GPT-3.5 is what they just released and their answer to GPT-4 is coming early next year

[@Slav4o911]

Considering how "good" is Bard I don't believe Google at all, I don't think their model will be better than GPT4.

[@gansogames4927]

Yeah I would love for openAI to have some real competition. The fact that a company like Google is lagging behind by an entire year is pretty disappointing.

[@PeterMcKeon]

The dominant force is always the first horse out of the gate, Google dropped the ball and won't be able to catch up.

[@LoopinFool]

Apple: Hold my beer...

[@medo_0x00]

@@LoopinFool tf you talking about. apple are still busy rereleasing phones from last year under a different name 💀💀

[@pdcx]

"Knowing a little less about your area, but more about the world in general makes you a better specialist" thats what AI is heading. training them just like real human learning the world instead of a restricted data set type.

[@mohammadmollaee3982]

Another mind-blowing video, another historical event, and another reason for INSOMNIA

[@bergrugu]

I swear yesterday when the news came out my first thought was "I need twominutespaper" love you

[@Psyopcyclops]

You picked the perfect subject to specialise in! AI is really blowing up, just as you’re maturing in your own education. Great job!

[@ChristopherCopeland]

9:00 “A jack of all trades is a master of none, but oftentimes better than a master of one”

[@ashutoshpatidar3288]

What a time to be alive!

[@unorubbertoe]

Plot twist.... 2 minute papers has been AI this whole. What a time to be alive

[@vladimirdyuzhev]

Then "What a crappy time to be alive!" feels more appropriate.

[@pv621]

​@@vladimirdyuzhevwhy are you scared of me

[@alexmarvin3093]

indeed, what a time to be alive!

[@arothmanmusic]

This is amazing. We may reach a point in my lifetime where both school and employment become unnecessary and/or impossible. What a time to be alive!

[@matowakan]

idk about school, education is very important

[@AIWRLDOFFICIAL]

​@@matowakanyou get the education from the LLM and or AGI/ASI

[@FRareDom]

It's a very interesting thing to think about when it comes to the future with ai

[@matowakan]

@@AIWRLDOFFICIAL You know at the rate things are going, I wouldn't be surprised if teachers are replaced with a child friendly AI, like what we seen in movies and games

[@sztigirigi]

Exactly. People are not as nice as GPT4 is.

[@GazFaaz]

Amazing! but also very scary. I remember The Animatrix, one of my favorite movie. Humans create program that can write a program and machines that can create another machine.

[@patrickegan8866]

That's an 11/10 level excitement "what a time to be alive"! I agree

[@agastyagoel6185]

For those not familiar with competitive programming, the problem shown at 0:45 is extremely difficult.

[@ertantosangcomandanterting1108]

Google's Gemini AI demo video isn't all real! It was found to be edited with no real-time voice interaction. Google has admitted to these edits, raising concerns about the readiness of Gemini AI for public use.

[@igoromelchenko3482]

I'm waiting for AI philosopher. But not a light-minded philosopher, but a hardcore one like Wittgenstein or Seneca or Kierkegaard 😊

[@tonyhere7004]

I would love something like this too, good sparring partner. Seems unlikely with the recent habit of hardcoding the morals of the company into it though.. 😢

[@vladimirdyuzhev]

Remember that LLMs are only probability networks. They can reshuffle what other people think and even make it more clear and concise and easier to understand but unlikely create something new.

[@igoromelchenko3482]

@@vladimirdyuzhev that's the saddest part

[@johnflux1]

@@vladimirdyuzhev I challenge you to even define what "something new" means. Like, give a specific example.

[@thomas.thomas]

A real person can go into the real world and see something new. Like a new situation. And then from that make new statements that no one ever uttered. An ai can only hear what other people already said or showed them. Ai can not see, hear, or read anything other than things that already have been said, have been seen and have been read. A person can observe the rain falling down on their window daily and look at a new pattern everytime, an ai can only see what someone else has filmed for them @@johnflux1

[@SinaMorovati]

You're into Petrucci and by proxy most probably Dream Theater and prog metal? What a time to be alive! Rock on dude!🤘🏻

[@TwoMinutePapers]

Long time fan!

[@clay8256]

I like seeing Google getting into the AI game more and more and also publicly release some tools more.

[@unknowjlm]

I don't : it's Google,remember...

[@maythesciencebewithyou]

google has been at it for longer than the others. They were just more reluctant with public releases of image generators and LLMs.

[@uku4171]

​@@maythesciencebewithyouhow long?

[@IanGrams]

@@uku4171 When it comes to language models, GPT-1 came out in 2018 and was built upon the transformer architecture that Google researchers developed in 2017. But before transformers there were other methods of Natural Language Processing (NLP) including word2vec developed by Google researchers in 2013. They have a pretty sizable research division that do some really fascinating work but unless it makes it to a product people rarely hear about it. Allegedly, Google was caught off guard by ChatGPT/Bing Chat because they didn't think LLMs were suitable for a shipping product yet due to their tendency to "hallucinate" or just make things up because the prompt asked for it. The early reports of Bing Chat telling someone it watched MS employees through their webcams was a pretty stark example of how they were right to be cautious.

[@deadpianist7494]

@@uku4171 they had this Gemini before chatgpt, they just didn't released it to the public

[@Canzandridas]

I used to joke with "give it two years and see where we're at" a few months ago. Now I'm afraid what will come in two years

[@robertstevensii4018]

This. It's going where we thought it would, where we preached it would. It's doing the very thing we said even the best of us could never be ready for. And somehow it is still shocking. Buy more popcorn. That's all we can do.

[@walltime1]

you know ai is advancing fast when even 2 minute papers is lagging behind in ai news

[@LambdaCreates]

yeah i'm not surprised if gemini 2 comes out in just a week they're speedrunning

[@sudhakarnayak1210]

And people still believe ai superintelligence is decades away maybe even a century away, which is as horrific as what it can and may do.

[@martiddy]

​@@sudhakarnayak1210I would say 2 years until AGI and maybe 6 years until ASI.

[@thomas.thomas]

the ai won't become "smarter" than the data it got fed tho - currently to me it seems like a talking encyclopedia @@sudhakarnayak1210

[@schumzy]

It's known that the MMLU paper questions has error - about 2% of the questions are wrong, so Google claiming their performance to 1 decimal point is a little misleading, as there is a built in error of 2% - so well within error

[@memesfromtheforsakenworlwi9218]

Just tried it, had a technical inconvenience for few months now that I couldn't really tackle fixing due to not knowing enough about it, bard gave the solution

[@bernard2735]

It is always a joy to watch your videos. Thanks for sharing your thoughts on Gemini.

[@alexmarvin3093]

what a time to be alive!

[@TheSparrowLooksUp]

All of this is very cool, but Dream Theatre is one of the most criminally underrated prog metal bands in the universe and John Petrucci is a true American Treasure.

[@SeanKula]

Wut

[@pupperemeritus9189]

I think it's prog as a whole that is underrated. Dream theater are pretty highly rated and famous among prog circles.

[@roldanduarteholguin7102]

Export the Gemini, Q*, Azure, Power Apps, Copilot, Chat GPT, Revit, Plant 3D, Civil 3D, Inventor, ENGI file of the Building or Refinery to Excel, prepare Budget 1 and export it to COBRA. Prepare Budget 2 and export it to Microsoft Project. Solve the problems of Overallocated Resources, Planning Problems, prepare the Budget 3 with which the construction of the Building or the Refinery is going to be quoted.

[@brandonacree4605]

I'm not sure if I'm alone in thinking this, but I'm not convinced that anything I develop with this tool, can't be claimed by Google. I've tried to ask it and it always hedges and says it can't promise anything. Kind of makes it useless as a tool if my hammer (or it's manufacturer) gets to own part of the house I built with it :/

[@GraveUypo]

yeah. completely agree

[@Megneous]

Asking an LLM legal advice is pointless. They hallucinate. They're very complex text completion bots, not lawyers. Consult your lawyer like a normal person.

[@thomas.thomas]

oftentimes the trained data wasn't theirs to begin with and the ai is infringing on copyright laws themselves

[@QuestionMan]

I could hardly wait for this video ever since I read the news about Gemini.

[@marktero]

Nice to see another fan of Petrucci 🤟

[@MikkoRantalainen]

It's good to remember that this is Google telling that their non-available AI can perform slightly better than the publicly available product. Let's see the real comparision when the Gemini Ultra is publicly available.

[@JackCrossSama]

A 24 hour friend that can teach you anything and work on your projects.

[@czerskip]

None of any existing tools can do that, sadly. Unless you already understand the topic really well yourself, you can't trust anything any "AI" says

[@dec13666]

_"What a time to be alive!"_ - Gemini, 2024 😐🤐🤖

[@joonglegamer9898]

One question to you as a youtuber. Often when we see press-releases like the one you're doing your own commentary on, why? There's already 100+ of this same video from Gemini, with each famous youtuber doing their own "side-explanation" or "narrating" to the very same video, explaining exactly the same things, what is the point?

[@curb_shifter]

asked it to summarize this video. what a time to be alive! KZbin Here is a summary of the video Gemini: A New AI from Google DeepMind Gemini is a new AI from Google DeepMind that is similar to ChatGPT. It is a multimodel AI that can do many things, including helping with math homework, summarizing research papers, writing code, and answering questions about a variety of topics. It is available in three sizes: Ultra, Pro, and Nano. The video also discusses the training of Gemini, which took place across multiple data centers. Finally, the video discusses the philosophical implications of Gemini, which is a universal AI that can beat specialist AIs at their own game.

[@haroldpierre1726]

I recommend you do some comparison testing between Gemini Pro, Claude, and GPT-4. Based on my own testing, Gemini Pro has been very disappointing. I tested prompts on all three systems - Bard with Gemini Pro, Claude, and GPT-4. Of the three, Claude consistently provides me with the highest quality content, while Gemini Pro generates the worst results.

[@egalanos]

My limited experience with Gemini Pro has also been disappointing compared to responses I get from GPT-4. It's a big improvement over PaLM2 though. Looking forward to trying Gemini Ultra. They're going to need to offer some trial period to use it as anyone paying for ChatGPT Plus won't be enticed by what is currently shown with Gemini Pro.

[@DreckbobBratpfanne]

Note: Bard seems to not use it for all users all the time yet and Gemini-Pro is just above chatGPT levels but below GPT4 and Claude-2 (and Inflection-2). Only the Ultra Version is a rival to GPT4 but hasn't been released yet.

[@someghosts]

@@DreckbobBratpfanneand won’t be until ‘early next year’, yet all the exciting features are ultra exclusive.

[@xr.melissa]

This is incredible, obviously. But with great power comes great responsibility. We already have a society (in the US) where 7th graders are barely at a 4th grade level and are struggling to read, let alone learn. So I'm glad they're being careful with who they roll this out to because society is more than likely going to get dumber the more we rely primarily on AI. AI is intended to be a tool, not a replacement for human intellect.

[@MrSchweppes]

We have to change the education system entirely. And the job market. In 1-3 years it will be obvious to everyone.

[@hosstam]

I agree, but do not underestimate the youth!

[@hrihori_art]

@@MrSchweppes I wonder what governments will do to create a new labor system or chaos awaits us

[@sea0920]

Maybe like Matrix, we can hook up cable to our brain and download learning materials from AI.

[@numoru]

Same issue persist human sensor input too weak to comprehend greatness. Impliment sensory enhancers in a great way to make sensory output reception greater. Hypnosis loop-Will signing out

[@tmendoza6]

I haven't had any luck with Bard so far in converting handwritten math notes, and I constructed the problem intending for Bard to do the conversion.

[@eugenesanderson6742]

This video posted by google is not accurate and has been heavily edited. It's an inflation on its capabilities. Sorry

[@MrMichaelLundberg]

I have tested it in Bard. First I asked if it used Gemini, and it said yes. Then I continued. I am not impressed. It could not answer a simple question like list the municipalities in Sweden. It can tell there are 290, but mostly only list 55, some times a few more. I tried several times and asked also to get it in excel format. It just cannot do it. And for some odd reason it blames it on that it only include the municipalities with more than 32% tax (I have not asked for that) and then it tries to correct itself, with the same result. Why I asked this question is because I have asked ChatGPT, Bing, Bard earlier and none of them could answer this simple question. So I wanted to see if there was any improvement. It was not.

[@tuseroni6085]

personally i like the idea of ai teachers. there aren't enough teachers, and certainly not enough for every student to have a personal tutor so a teacher that can help a student to learn, make questions, judge answers, and help get to know them and motivate them to learn is, to me, a great thing. i worry a bit about who controls the curriculum but i worry about that with public schools too.

[@GNARGNARHEAD]

John Petrucci mention, oh you better believe that's a thumbs up

[@Crayoness]

gotta let the competition keep going

[@csilval18]

Terminal velocity is such a banger album. Petrucci's the GOAT

[@PuppetMasterdaath144]

Ive understood that when it comes to inference correlation is the best methodology to attain "facts" on certain topics that are not deductable

[@plasmarade]

I fear frightened for my job security, and im only in my high school years...

[@fernandodutra3788]

First thing I thought when hearing the name of this AI was indeed John Petrucci’s song😅 great to see you are also a fan!

[@jendabekCZ]

As expected, the video has been staged, and it can't really do what has been presented in realtime.

[@crimsonstar108]

The "buzzes" from this digital Karoy's voice are fucking hilarious It reminds me of Kate Reading reading Pattern, which is funny because my ideal AI would be modeled after a spren

[@virajsheth8417]

"AI explained" covers in more depth and nuance. I love two minutes papers also.

[@torarinvik4920]

What a time to be alive!

[@wolfgangsanyer3544]

frickin love this channel

[@faqeerhasnain]

IMPORTANT : For those utilizing GPT-4, they are still engaging with the latest technology. Although Gemini is available, only its Pro version is currently on the market, not the Ultra version, which surpasses GPT in capabilities. The Ultra version is not expected to be released until next year, though the exact date is not yet known.

[@galvinvoltag]

AI is getting closer and closer to outsmart us. Maybe the little AI will be able to end us soon, I'd be so proud. What a time to be alive!!!

[@ТуанНгуен-ь5п]

Gemini is this good because of that new DOJO supercomputer, also, expect something amazing from NVIDIA because of their new specially designed neural learning graphics card

[@zerorusher]

Gemini seems to have a very similar performance to GPT4, which is impressive but still behind openAI which is possibly working on a far more advanced model. The one question that's on my mind, however, is how fast is inference time and how much API access will cost. GPT4 can solve a lot of problems using techiniques like CoT and ToT and Autogen, but until inference gets a lot cheaper, only those with deep pockets will be able to implement that.

[@Slav4o911]

GPT4 can write a whole game all by itself, with code with the dialog inside and what not, but it would not do it, because it's censored. If Google is going to implement the same limitations, their Gemini would not be better... and for GPT4 to compete, open AI will just have to turn down the censorship a little bit. There is a whole topic on Reddit about how GPT4 is censored to not do your programming work for you, and that after a time it will become less helpful.

[@sahinyasar9119]

We are truely blessed to experience this age of change. Literally what a time to be alive.

[@kinngrimm]

Not much more and coders can say goodby to their jobs. Software architects are closely behind. The best of the best, Highlander style, may still get one. Then again, if let go, said coders and architects can create their own company just as easily and become the new competitors to those that just kicked them out. This might get very interesting indeed. Should we all start to safe up for our own supercomputer racks?

[@thomas.thomas]

do you want to invest 20k into a supercomputer that after 5 years needs to be replaced by another 20k one?

[@kinngrimm]

@@thomas.thomas Towards that issue was my last sentence direcrected. We already have concentration of wealth in power that many claim to be anti-democratic, maybe not in its intention but in its effects. There are studies out since 2010 about that, where it was shown that rougly 130 companies own 95% of all companies world wide. Due to shareholding, alliances and even marriages you can break that down into 5-30ish conglomarats depending the metrics one uses. That was 2010 stated by a study by swiss university. I don't assume that to have gone any better since, quite the opposite. As an unexpected outcome, the scientists back then said this concentration of wealth would be the biggest threat to democracy. Now to your statement and my hint, as that is not the end of it. What is the one thing companies activly try to avoid? Costs. Which costs are usually the once that are highest and under certain circumstances the easiest to reduce? labor costs, correct. Now while it is true that currently ChatGPT and other AIs are rather enhancing productivity people in their work, depending on branch i would assume that hiring will fall flat, they may stick with their original staff, but due to higher workout put, but not necessarily more contracts to do more work ... . You see where i am going with this. Add robots like the once from Boston Dynamics into the mix and i am not really seeing a future where people are doing a lot of labour. The one point the owners of production lines will have is to gurantee that there are still paying costumers, where i would guess something like a general basic income would come into play, not matter how capitalist some may think of themselves, if noone has money to buy their crap, they will rethink their positions. Now letting aside how exactly such a system in terms of money flow would work. AI will increase market pressures, those who already use them against those who do not. Which brings us to the beginning again, the concentration of wealth ever more increasing. Now i don't know if we end up like that, it seems rather likely to me, that we are in process of creating a new monarchic system, only that it is just based on inherited money in the end as at some point there is hardly a chance left that some disruptive tech may still make a dent. Till then though, there is still a lot that will change.

[@franciscomagalhaes7457]

I think this might be a big turning point. It's multi-modal and it responds well to real world scenarios, or maybe better said, "human experience scenarios". With this sort of capability, making smaller, task-focused agents might be much easier... and start replacing a LOT of jobs in the next year or two. But even with that said, I'm still a believer in the benefits of AI. I still can't bring myself to believe it, but I keep reminding myself: technological breakthroughs do exist, and maybe we're alive to see one of those that will reshape the world forever. What a time to be alive...

[@autohmae]

what we have now is already doing that. look at Alp;hafold and how much it's being used

[@franciscomagalhaes7457]

@@autohmae those are specialized neural networks. When I say "turning point" I'm talking about the pursuit of AGI.

[@autohmae]

@@franciscomagalhaes7457 I'm just saying, we don't need AGI for huge impacts on society. In theory we could put pretty much everyone out of a job without AGI, just more specialized ML (ML is machine learning, I don't really like the AI acronym, because it's not really intelligent what we have now).

[@franciscomagalhaes7457]

@@autohmae You can use AI as an umbrella term for everything that at least *attempts* to emulate intelligence, or at least the capability to do something without being specifically programmed to do it. And yes, I think you make a good point, but I also think the time and resources it takes to train specialized AI for a variety of tasks, from home plumbing to civil engineering to call center support agents, is incredibly high when compared to a system that can take in information in a more generalized fashion. You're not wrong, and I'm convinced that as it stands, LLMs are already well on their way to turning the world as we know it upside down, but I think AGI would really compound that and make it happen in a matter of a few years instead of decades.

[@fiffy6572]

this is amazing, an amazing learning tool for sure!

[@Exilum]

Now let's see google make geminis to compete with GPTs. That's the most compelling form of distribution in my book.

[@PauloSamurai]

Petrucci is amazing!! ♪

[@thethoeby]

Do you have pre-made audio clips and they got mixed up at 1:36 ? If so you might want to check it there was some sort of mixup, I think I remember hearing it in yesterdays video too.

[@TwoMinutePapers]

Nope, all of them are freshly recorded. Always! This one is from today with me being a bit under the weather. 😀

[@21EC]

I'm glad there's such a great AI that is called like the astrological sign I have, which is Gemini :) who else here is a Gemini?

[@DennisHaney]

I tried it with the last handful of tasks I used chatgpt4 for in the last couple of days. Task 1, was to help me design some slides, and it did slightly better (pretty much same output, but the image was nicer). Task 2 was programming, and it failed horribly, even 10 prompts in, the code generated did not fulfill the basic requirements given. Task 3 was programming and it did similar, but required much larger prompt than chatgpt. Task 4 was some silly discussion about cats, and it did similar. Task 5, generate a picture, it failed horrible. So I will stick with chatgpt for now.

[@autohmae]

Supposedly this is Gemini Pro, and Gemini Ultra isn't available yet...

[@Luid101Clips]

Super human performance on a large array of tasks previously dominated by humans... is this AGI?? 😮

[@krabbe3]

The thing of AI/ AGI and the comparison to humans always lacks one important point imo. Our brains uses like 20 W to do all the crazy tasks and all that creative stuff and research and ... we are able to do. Gemini for example was trained in x data centers with thousands of gigabytes of data and god knows how much power this took (TW probably, I dont know). So in my opinion it is quite "normal" that AI outperforms humans at several tasks given how much more ressources are used by AI vs the human. I mean imagine our brain would have the processing power and storage capacity of several google data centers. How smart would this superhuman be. I really am stunned of AIs progress and I really want to live in this future because it offers so much possibilities. I mean for example in research you could focus on the research itself and let the AI do the boring stuff like writing the papers, writing simulations code of your ideas,... WHAT A TIME TO BE ALIVE!!! But if people compare human vs AI right now for me this feels a bit like comparing apples with pears. But I am very excited for the future and wanna see where this goes!

[@wycliffe_ndiba]

Could you make a video for coding with Bard because I still think ChatGPT and Copilot are better (or maybe I'm better attuned to prompting chatGPT than Bard) You could answer questions like are there different prompt structures between these two platforms and how can we best make the prompts (+ tricks like "let's think step by step" and "let's take a deep breath")

[@simonprovencher6007]

what a time to be alive!!!

[@julienmans3359]

John Petrucci??? I see you're a man of culture as well

[@tanakablack5271]

i was waiting for this video

[@DarkMuzishn]

Petrucciiii🙌

[@TwoMinutePapers]

🙌🎸

[@JohnPhrey]

People Going to be replaced by Ai : "What a time to be Alive!"

[@tycse]

Ayyyy shoutout Petrucci!

[@sirpsionics]

It's a shame you don't have more subscribers

[@blzrdphoto]

My bard says it’s using Gemini pro! Woohooooooo!!!!

[@berkeokur99]

Another Petrucci fan!

[@Chessmasteroo]

It was reported that the duck multi modal video was misleading viewers to believe it was real time. Google stated in the fine print that it was edited for brevity and certain frames were edited. Still impressive just a little false impression.

[@willmurrin9344]

John Petrucci, good shit.

[@MrSchweppes]

We have to change the education system entirely. And the job market. In 1-3 years it will be obvious to everyone.

[@himanshuthakur4171]

Meanwhile Sam Altman laughing in the corner waiting to introduce Q*.

[@Chris.Davies]

It's funny isn't it: We are all living in the future! It's just that the future looks a lot like the past. WHAT A TIME TO BE CURIOUS!!

[@helmutzollner5496]

Great program. Thanks

[@StephenRansom47]

😮 I shave dreaming of such a thing since I first saw R2-D2 and 3PO in 1977, since I saw HAL in 1975. I really can’t wait for this to be ubiquitous. You Hear Me Santa! 😊

[@alexmarvin3093]

you... shave?

[@zorayanuthar9289]

Then you'll have a "I'm sorry Dave I'm afraid I can't do that..." 😂

[@StephenRansom47]

@@alexmarvin3093 now, I’m not going to edit it … lol

[@StephenRansom47]

@@zorayanuthar9289 😂 Then I’ll be like … you lie. You can do anything. Now open those doors, or I’ll have R2 beat you up. 😤

[@manooxi327]

Thank you ❤

[@JackBond1234]

It's a shame so much technology is wasted in the hands of a malicious company

[@tricktap8502]

Sometimes I remember how insanely smart people for creating these systems.

[@chasewatkins3096]

2:50 Anyone else hear the crazy warble in his voice? Is this channel narrated by an AI voice now?? I've been wondering every since I started watching again a week or so ago.

[@cybercoltz9054]

He's been scripting himself since 2021 and it's extremely disjointed and weird

[@shawnfromportland]

what a time to be alive

[@lemontec]

„Getting this right might be an existential issue“ An existential issue for humanity, right? RIGHT?

[@ZOMBY2511]

The showcase video is fake, is nowhere near that level.

[@Nonkel_Jef]

Hell yeah, Petrucci

[@jeremiahwat1]

Makes me wonder if with an internet connection, you can make it connect to openpilot software and have address to address selfdriving capabilities.

[@cicada-ft6eb]

And my stupid ass, gemni : can you give me deez nuts joke