My guy was waiting in the corner thinking “I’m gonna be better than them one day”

I've found Satya Nadella's KZfaq account! :)

Or any other rival company, so we could at least say “the Chandler/Bing war”

PIVOT!!

MICROSOFT PLEASE MAKE THIS A REALITY!

I know, right? don't they have anyone over there who knows HTML? HAHA

They have to fix it first as after testing bard, gpt 4 and chat GPT myself with different logic puzzles GPT 4 passed 9 of 10 (it can not solve the 2-4-8 riddle no matter how much info I give it for those unfamiliar you need to give up your positivity bias to solve this riddle) chat GPT passed 6 of 10 (because it could not keep track of completed riddles (also failed the 2-4-8 riddle) ) Bard however only solved 1 of 10 of the logic puzzles after some real effort on my part to tell it how to go about it. (admittedly bard is still in demo and I only had access because I had signed up so it may improve)

But who would use that if chatGPT is available. Bard would have to have to be insanely better in some way for people to choose to use it and tolerate ads. chatGPT subscription model seems much better right now.

@@erikburzinski8248 I probably get out performed by Bard 🥲

Excuse my utter ignorance but can I ask you if you know of a good source where I can learn more about this? I've been eager to learn more about how these AI systems work but I've either found overly simplistic articles or straight up complex research papers, not much in between.

If I were a billionaire Id be constructing a personal server room already. Cant wait for this tech to become affordable and less resource intensive

@@smryt9728 I'm probably the wrong person to ask--I studied computational linguistics. My favorite videos are @TwoMinutePapers though he focuses on graphics. It's a great overview a week or two after someone publishes something significant. I like @NerdyRodent and his install videos on youtube. Recently he made one on text-generation-webui which lets you run LLaMA and a bunch of other language models. Ilya Sutskever did a couple interviews which are quite approachable in the last month or two, but they're 1-2 hours each, so I watched them at double speed.

after testing bard, gpt 4 and chat GPT myself with different logic puzzles GPT 4 passed 9 of 10 (it can not solve the 2-4-8 riddle no matter how much info I give it for those unfamiliar you need to give up your positivity bias to solve this riddle) chat GPT passed 6 of 10 (because it could not keep track of completed riddles (also failed the 2-4-8 riddle) ) Bard however only solved 1 of 10 of the logic puzzles after some real effort on my part to tell it how to go about it. (admittedly bard is still in demo and I only had access because I had signed up so it may improve)

@@erikburzinski8248 Hm. Google doesn't have a working product. Their AI dominance was a bluff.

@@mikicerise6250 they are working on PalmE which has more then enough data for a huge jump however it will take quite awhile to get something to work with it and I am pretty sure bard will be moved over to palm E eventually which will allow it to exceed gpt 4 significantly however its current state is all I can test

@Sam Attenborough just retested it and unlike both gpt 3 and 4 which were at least able to understand the test even if they couldn't pass Bard could not even understand the test

At this point Google supporters look like they're coping, "they have better models just wait and see"

Agreed, I think Bard is running a relatively light model compared to ChatGPT or Bing

Social and ethical concerns, LMFAO. If they have this, it’s not being released because it cuts into their search business and when it isn’t good enough to do that, hurts their brand image. As if it’s about anything else than money and market share 😂

Absolutely not

Which powerful LLM are you referring to? This is pure speculation.

Google announced that they would soon use their more advanced model PaLM to power Bard two days ago.

yes it is personally I currently am betting on 6 months or sooner before the singularity (100$ with my best friend) what so you think (requirement for singularity is an AI improving itself faster then humans can comprehend however sentience is not required)

On an Ai video

what? since when? I'm using it as we speak.

I hope it was a glitch, but have been having trouble with it rewriting html and css into a razor pages format (not sure why as I don’t believe this would be that hard , and I’ve seen it do way more impressive things)

It's not that great. Chstgpt beats both bard and bing at the moment

I'm using both now, and I never even touched Bing.

I switched to bing and edge over it. Both are amazing

Dawg lambda was trained before chat gpt, and entire company is built off the transfomer model which I'd a paper by google

@Sam Attenborough Lingering trauma from Windows 98 ;)

I can only speak for myself, but practically speaking, I've already substituted Google search with Bing Chat search. It's way faster, very accurate, and generally more efficient. I now use the "new bing" 9 out of the 10 times I would have used Google.

I don't think so. They seem like two different scenarios. The AI arms race is a matter of working on an emerging technology, which arguably has sourced some of its training material in questionable ways. SVB, well that's just another bank failure, I'd argue that it's nothing all that special.

It would be very unlikely. Microsoft owns 49% of openai which would make any deal extremely hard and while Microsoft doesn’t have the profit deal for gpt-4 they still are the owner of 49% of all shares and would still benefit from its success. Also considering googles efforts in Ai it would make little sense to buy out something like that after putting in all that work. I hope this helps

@@Ok.ok. oh i thought Microsoft owned 49% of ChatGPT not OpenAi whoops lol

It is a easy enough mistake to make honestly

after testing bard, gpt 4 and chat GPT myself with different logic puzzles GPT 4 passed 9 of 10 (it can not solve the 2-4-8 riddle no matter how much info I give it for those unfamiliar you need to give up your positivity bias to solve this riddle) chat GPT passed 6 of 10 (because it could not keep track of completed riddles (also failed the 2-4-8 riddle) ) Bard however only solved 1 of 10 of the logic puzzles after some real effort on my part to tell it how to go about it. (admittedly bard is still in demo and I only had access because I had signed up so it may improve)

It's bullshit haha

Great video btw Jack 😊

Google has since 2019 been integrating AI into the search itself. The difference with gpt4 and Bard is that they can use a natural language model to split combined searches and recombine them like a human would eg found me all capital cities with a population over 10M in any countries with a GDP under $10K. A regular search can’t do that unless it has already been done on the web somewhere while the AI can combine the results from the two searches.

First off, to be clear, purely for searching subjective & not so obvious information on the Web, the traditional Google and Bing are still the way to go as you can vet the factual correctness & bias a bit easily based on the source reputation. But, let's say, you have a question that is not subjective & just needs a quick answer: "Which carrier is the cheapest to mail a 10lbs box within the USA?" Instead of you going through 10s of links, based off of the training data (or) web search on the backend, GPT summarizes for you what it understood. Basically, automating that process is what this AI can do. The problem however is, if that information is not present, GPT just gives out something it thinks you will believe is true (as it is trained to be a helpful assistant, even if you're asking an impossible question). It does not know it is lying and making up stuff as it cannot reason and is not self-aware YET! So, always double-check with a quick search if it is something important.

I've been using new bing since the day It came out, and can offer my perspective. It is a more direct interaction, where you can just ask a question. Let's say you have to study Martian polar ice caps for a project. Here's what I would do on google: 1. Google "Mars" 2. Open Wikipedia 3. Read the article 4. Realize that it only contains 2 paragraphs about the polar ice caps, and that one of those 2 is just about how they're made of water 5. Go to the Wikipedia for the polar ice caps 6. Get good info, but realize that you aren't allowed to cite Wikipedia 7. Go through wikipedia's sources By the end of it, that's an hour or more gone. Here's the bing ai process: 1. Bing "Mars" 2: Click on "Chat" 3: Bing summarizes some facts about mars, and asks "Is there anything specific you would like to know about Mars?" (direct quote) 4: Ask it "Can you give me some good info about the polar ice caps that would be good for my project?" 5. It gives a helpful list of facts (a lot easier to read than an entire Wikipedia article) and cites all sources Overall, I find the AI best for complex problems ("Will the large blahaj from Ikea fit in the back of my 2003 honda civic") and questions where you'll want a lot of clarifying info ("What are some facts on Mars?") I've also found it useful for my writing, as I can ask it for ideas, pick the ones I like, have it clarify more, change some parts, and then write those into my story. Feel free to ask more if you want! I've found the AI quite useful in my day-to-day life.

Id have agreed with you a year ago. Now Bing is just better. I have no loyalty to any brand. I would use Yahoo if they came out with an AI augmented search even lol

@@Shellll yes I agree. Although I haven't used Bing, chat gpt or Baird. They have come so much further. That's why I think they shouldn't be comparing them like they have. Google is a search engine company first and foremost, Bing is like a little side project for Microsoft and they are still getting ahead of Google in tech. I don't have loyalty either just that my phone has Google preloaded so it's what is used. If google loses the deals with android phones, it would probably collapse with Apple having safari.

probably hundreds of thousands of times more power

the singularity has happened, sort of. atleast in terms of online jobs, everything is already replacable, even entrepreneurs

we are to late the genie is out of the bottle as now that it is international for AI development any country that tries to implement safety will end up too far behind a country that hasn't these needed to be established globally last year

yeah because this has the ability to make us redundant

No Microsoft kept their office of AI responsibility and broke up the 'ethics' activists.

False. Microsoft reduced their team from 30 to 7 in October 2022, maybe because it's all outsourced to OpenAI.

Human-Feedback! When you downvote a response, ChatGPT understands it is not helpful to you, and tweaks itself, if allowed to. It is possible that a few folks at OpenAI are involved in this tweaking process, but yeah, they're training the model further on our conversations.

Not directly, but OpenAI does use the interactions to fine tune the chatbot. In particular, many of us converse with ChatGPT in ways that intrinsically help make it more rational and accurate, and OpenAI probably knows who we are. ;) 4Channers, this time around, will just be flushed down the toilet.

@@wolfie1001 I get the voting, but not so sure about general conversation. I think Sam Altman said they have some threshold for content quality. For example reddit posts with at least one upvote etc.

what Elon needed to do was put more hardware in the cars like radar as the more data an AI has the better it can perform even if that data is useless to a human

Google has been on the cutting edge of AI for years.

@@ASLUHLUHCE so was Kodak with digital photography. They just never released to the public.

@@38josue91 Oh lol, a Kodak engineer invented the first digital camera

also after testing bard, gpt 4 and chat GPT myself with different logic puzzles GPT 4 passed 9 of 10 (it can not solve the 2-4-8 riddle no matter how much info I give it for those unfamiliar you need to give up your positivity bias to solve this riddle) chat GPT passed 6 of 10 (because it could not keep track of completed riddles (also failed the 2-4-8 riddle) ) Bard however only solved 1 of 10 of the logic puzzles after some real effort on my part to tell it how to go about it. (admittedly bard is still in demo and I only had access because I had signed up so it may improve)

and just wait until microsoft launch a new windows OS based around GPT.

Edge has actually become decent lately, especially compared to Chrome.

@@WG55 Yes, but there are other reasons than that not to choose Edge.

@@varnull6120 I actually started using Edge (also hated it with passion, before) - because of their bing plugin that reads and answers questions for you. Man, I used it for this test (not mine) and just asked it: what's the answer for question 1, 2, 3 and 4? and it went boom. Its accuracy is a little bit overhyped, it got some answers wrong, but "surfing" the net for info seems to be a thing of the past. Especially if you know what you're looking for gonna take time to find.

​@@varnull6120 What are the other reasons?

@@AB-wf8ek For me it comes down to the breaking of adblock in chrome, which Edge will inherit, and just personal preference. I like firefox. I shouldn't need any other reason. I use edge professionally, it works great, but it's my choice to switch or not - and the aggressiveness with which MS pushes edge leaves a bad taste in my mouth.

Its all in the prompts. The trick is including phrases like: "Lets think step-by-step" or "Be sure to iterate over your answer to verify its correctness before responding to me" - - these are basic examples but there are signifcantly better ways already

@@Shellll okay, sure. I’ll grant you that. Now think about your standard internet user. Do you really think someone’s grandma or racist uncle will know about the prompt tricks? Or will they just search “how should I invest my pension?” and then believe the search bot when it says something obviously ridiculous or dangerous?

@@juliegolick oh yeah I agree with you on that. Its up to Microsoft(referring to Bing chat) to bake in those sorts of prompts to minimize misinformation for novice users

@@juliegolick if you think about it, this has been the same with "the internet". haha. People still read stuff and take it at face value. No AI needed there. And if I had a dollar for every time I did the same...

@@mobbs8229 Oh, agreed. Absolutely. But adding AI search bots to the mix will just make it WORSE.

Google is right to worry that the combination of bad press / commentary (tech 'journalists' from the Web3 hype cycle tribally fixating on OpenAI GPT for being the first doing a lot of unnecessary damage) could see GPT grow faster through being used and therefore user in loop trained orders of magnitude more

Competition is good.

@@cgrw1 after testing bard, gpt 4 and chat GPT myself with different logic puzzles GPT 4 passed 9 of 10 (it can not solve the 2-4-8 riddle no matter how much info I give it for those unfamiliar you need to give up your positivity bias to solve this riddle) chat GPT passed 6 of 10 (because it could not keep track of completed riddles (also failed the 2-4-8 riddle) ) Bard however only solved 1 of 10 of the logic puzzles after some real effort on my part to tell it how to go about it. (admittedly bard is still in demo and I only had access because I had signed up so it may improve)

yes, but once in a decade or so, there's this technology that's *actually* worth the hype. The internet in early 2000s, smartphones in 2010s, and now this...or?

@@mobbs8229 some things do work, but cellphones contributed to cyberdependance a lot, so people are stasrting to slowly move away from the over usage and are taking step to make their info less profitable. As for the internet nowadays yes its a revolution but that too has its up and down, lots of disinformation on that(you dont have to believe me), lots of new social probleme (online bullying, arrassing, stalking.....). I recommend the Adam Conover video for more information on how AI is already a failure.

"My god, Duncan climbing a cliff is so hot that I'll literally come watching it from below" -Fremen woman, God Emperor of Dune by Frank Herbert

Cringe

I'm tired of the work of fiction being used to explain the real world

however after testing bard, gpt 4 and chat GPT myself with different logic puzzles GPT 4 passed 9 of 10 (it can not solve the 2-4-8 riddle no matter how much info I give it for those unfamiliar you need to give up your positivity bias to solve this riddle) chat GPT passed 6 of 10 (because it could not keep track of completed riddles (also failed the 2-4-8 riddle) ) Bard however only solved 1 of 10 of the logic puzzles after some real effort on my part to tell it how to go about it. (admittedly bard is still in demo and I only had access because I had signed up so it may improve)

Ikr, the swearing doesn't fit the vibe of old school TLDR

Its more about appeasing KZfaq algorithm