The Slow Death of Google's AI
Вставка
- Опубліковано 4 чер 2024
- Join My Private Community - / theaigrid
🐤 Follow Me on Twitter / theaigrid
🌐 Checkout My website - theaigrid.com/
Links From Todays Video:
Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos.
Was there anything i missed?
(For Business Enquiries) contact@theaigrid.com
#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience - Наука та технологія
The best part of AI is that it can summarize your videos.
😂 really really shocking
I wish
@@kristofszabo9800eightify chrome extension, just make sure to save your free usage for this channel 😂
Lots of opinion, lots of "thatss craazzzy" no substance just random opinions. I've heard better rants down the pub
6 min in and I realize I wont get anything from this... Shocking
Let’s not forget how Google had years to start the AI war but they waited and waited until OpenAI dropped ChatGPT
they overlooked the value of AI.
Google invented and patented the Transformer architecture in 2017. This architecture is used by all cloud AI companies. They also patented conversational models, which are the basis of all modern chatbots. They have been demoing this stuff at Google I/O for a number of years. OpenAI's innovation was to wrap an ultra-simple web app with a chat UI that was dead simple for non-tech people. Kudos to them. Imo, Google has product management issues, and OpenAI has company stability issues. Both, not that great for enterprises interested in GenAI.
OpenAI it's a centralized company is not free, it's has it's own political agenda, future AI has to be decentralized and not controlled by companies but by the people like Bitcoin, or cryptos, so in the meantime all people we are just sheep behind a company that manipulates in almost all responses!
@@lance3401 I agreed with you up until you said Bitcoin or cryptos... which are artificial commodities from overpaid fintech frat boys, essentially no different than pork bellies.
Googles mentality is always " we have no intention of releasing this product to the public".
That's because it's become essentially a component of the MIC.
Google died the day they removed the ‘don’t be evil’ motto from their existence. Everything since has just been the corpse rotting.
the fact that google since the mid 2010's has been showing us these amazing AI tech demos then when forced to actually deliver something it's laughably bad, just proves to me all their demos have been completely faked to juice investors.
Google lost the AI race.. when they were racist with their search results.
The "AI race" (whatever that means) will continue for decades. We are just getting started in finding good use cases for most enterprises.
Google has been on the decline for ages since it started prioritising politics and concentrating less on innovation
@@soggybiscuit6098 All large enterprises have political strategies. Google is no different. This is more an issue with the US's form of capitalism rather than an issue with Google. Btw, Google's stock is up 250% over the past 5 years... not exactly declining. They have 112K patents, including Transformers, which is the architecture that OpenAI uses. Google has a lot of issues with product management and marketing, but I wouldn't say they concentrate less on innovation.
@@frankgreco mUh CaPiTaLiZm Is BaD!!1
No. Google has not innovated in two decades. Everything they have they've PURCHASED since then. Pro-top: when the dollar loses valve, stock prices go up.
Every major company in history has failed from complacency. When they focus on bread and butter products with blinders on thinking the gravy train will run forever.
Google just love to push out half baked stuff then cancel it.
Google is the next IBM
Google AI is lame. Meta's V-JEPAs are boring promises. Microsoft is the only one winning being partners with OpenAI. And Nvidia of course.
TLDR; Google execs are busy theory crafting, “hey bro, wouldn’t it be cool if we did this?”
Meanwhile, OpenAI is sitting back watching saying, “Yea, that would be cool… I’m going to go ahead & do that instead of daydreaming.”
Paralysis by analysis, I believe the phrase is.
Google is in bad shape. Everything they do, messes up, despite the fact they are the one who made ai possible 😅
Google image search is now worse than Yandex !
@@MilesBellas The problem with google is censorship.. I have gemini pro and sometimes it refuses to generate code because the prompt mentions a bomb. In this case it was instructed to write a game, so go figure. I had to tell it that it was a virtual bomb.
@@MilesBellas yandex??
@@MilesBellas yandex?
You just put into two sentences, what he is babbling about for 21 minutes.
Funny how they have all that computational power, some of the most brilliant minds and resources for years and for some reason, sometimes seems like people don't stop to think they are choosing to be the silent giant.
Google = Censorship
Nah... Google = Ads. Google search merely probabilistically reflects content on the web. Their ads are the real headache.
Everybody wants some form of censorship. Even you.
For anyone using GenAI in industry via APIs, friendly note that the new Gemini 1.5 Flash is 87x cheaper (compared to GPT-4 when it had first come out) and is very near the top of the benchmarks and above the early version of GPT-4. It actually performs really well. Great vision, extremely fast, good high rate limits and that unlocks incredible amounts of use cases that were cost prohibitive before.
Some people like to hate on Google, but at the end of the day, at the given prices and context length, it's just hands down better for many use cases today.
Thanks for adding more color to the story.
Sadly, the problem is the philosophy around AI these days.
In the past Google thrived because of initiatives like Tensorflow + Keras.
Sadly the new leadership's focus is too focused on immediate money making rather than advancing the field. There is no logic in terms of Claude beating Gemini, There is no logic how Llama is leading the open model race. Yes google's models are cheap but gpt 3.5 solves the overal majority of problems and is also very affordable.
Bad leadership is what is killing google. The immediate profit view is impeding innovation
How does it compare to 4o ?
@@36mcast hey, I love Llama Claude, the GPT models. They're all great for specific situations. But the prices speak for themselves: gpt-3.5T is 42% more expensive than Flash. Llama 3 70B is ~650% more expensive than Flash (via AWS for example). Even llama 3 8B is 14% more expensive.
Google is killing it if you think about it.
Claude's Haiku is ~30% indeed cheaper for inputs than Flash and that's great if your use case permits it. Anthropic are doing great work too!
If they didn’t outsource 99% of their jobs. They wouldn’t have this problem.
I knew I worked 3 years for various Google projects, but couldn't talk about it, or felt like it. Once I felt confident I helped train their AI, I proudly spoke out as "Well, I helped created the demon!". Once the partnership ended this past summer, I finally learned I was right, but also quite wrong, being the data was used for... Bard. Something not even worth mentioning as seen in this video, just like I stopped since, but thank you for doing so - haha.
I'm so against outsourcing, such a bad idea in most cases.
I think google at this pace, it’s becoming Nokia or KODAK. Goggle needs to restructure like their are dying.
Let's just call it what it is, Google's obsessive mission to be "Safe and Inclusive" and maintaining that image at all costs means that it is the MAIN priority. They have become single-mindedly obsessed with that goal that all else (like a good product) has become a backseat concern.
And they end up with something that is neither safe, nor truly inclusive with serious contradiction in its "morality".
Nah... it's more "Safe and Responsible"... let's face it, LLMs naturally "hallucinate"; that's precisely how they work. They select paths according to probabilistic distributions with no guarantee of "correctness" as with a search engine (where "correctness" is defined by returning the actual data). There are techniques to mitigate hallucinations, but they require more LLMs, more cross-LLM verification, etc. You can persuade any LLM to travel down an unexpected path, not just Gemini. We are in the early days of production-quality GenAI... very, very early. Google was forced to accelerate its use of LLMs thanks to Microsoft, which is "going fast and breaking things" at all costs in a competitive war with Google. Google's problem is not their engineering; they have premier tech talent. If anything, they are too engineering-centric. They have other business and product strategy issues, not to mention marketing.
I think that thoses are excuses for their low quality employees. In the context of a corporation, the progression is burocratic and likely is difficult to create a meritocracy. I think that they use the security consern to mask their low capacity at the management and technical level
@@tomaslanza8564 They have the same quality employees as their competitors. There's a lot of churn among the big 7 so the same players get their cards punched. And what big enterprise is not bureaucratic? They have world-class engineers, but their product management and marketing are sub-optimal. Their exec management is far from optimal.
8:20 just imagining someone at Apple HQ looking at that screen and having a clutter meltdown
i was really surprised recently when i asked google gemini advanced to summarize a youtube. the entire summary was a hallucination. i pasted the transcript into claude opus and chatgpt and gpt and got great summaries. it just surprised me that googe, who owns youtube, would have an ai chatbot that literally gave an entirely erroneous summary of one of their videos. makes me not trust gemini much.
It shouldn't be surprising to see the same tech to censor and restrict videos and we seen how bad they are at that.
Pichai is not a visionary.
He may be an illusionary. 😂
Like Steve at Microsoft and see where they are now after their winter. Maybe Google catches the next wave in a decade
Google needs to go dark for a year. Do whatever works. Disappear most of the crap that makes the search engine garbage. Make it good again. And just work on getting things to get it right & make stuff GOOD again. Focus on bringing 4 products to market 1 year from now. Whatever they are. Be AMAZING. 🤷🏻♂️
Edit - I found it funny Google engineers were leaving because they weren’t paying attention. - I thought attention was all you need. 😎
Just pointing out that iPhones are nowhere near dominant in the smartphone area.
Android has almost 71% of market share.
Minus china and it’s 50 50
it's not an hallucination, it copied that response from reddit
the truth is it does not take a big company to write good software. a small team of dedicated people who can align their goals and work together well will likely outperform all the big tech conglomerates. It's kind of how all the good things get started.
Look for the Engineer/HR ratio.
"The video isn't real"
translation:
Its fake.
Google didn't intend to give us AI.. thats the difference... they wanted AI... but just for them.
Plot twist: google achieved agi and its trolling them into feeding it more data, more compute and taking the guardrails off.😂
03:50 Company Size
The large number of employees can delay innovation
Comparison with smaller and more agile companies
07:01 Computing Capabilities
Google has advanced computing resources, but does not use them efficiently
11:00 AI Demonstrations
Criticisms of Google’s misleading AI product demonstrations
14:01 The Innovator’s Dilemma
Established companies fail by ignoring disruptive innovations
19:25 Focus on AI Agents
The next phase of Google’s AI may focus on agents that plan and act in the real world
thanks for this breakdown. Google has been wilding for a while, thanks for gathering up all the details in one place!
Google basically owns the gas and the parts while Open AI built a really fast car. OAI "Look how fast I can go!" Google looking at their resources "Let us know even you need a refill." Taking a peak under the hood all the while.
Massively improved spoken content, well done.
Please don’t be hoodwinked by this very smart company. They knew before any of us that most AI innovations are probably going to detract from and weaken their grip on the massive incredibly lucrative online advertising industry over which they have virtually monopolistic control. This probably also accounts for their apparent failures in AI, failures they knew would happen but that are subtly intended to discredit AI, again to protect and maintain their dominance in search advertising. Who would want to give up the world’s best business model just to become one of the million companies fighting to offer the most used, but hardly profitable, overhyped, underperforming, highly unreliable “word-guessing party tricks” that is the Ai on the streets today?
Google is the yahoo of the internet soon
0:03 Google Air?
Great analysis. Large companies can become so large so as to be unable to pivot. Two classic examples: Xerox invented the mouse/GUI - Steve Jobs took it from there; Kodak invented to digital camera - and then shrank to its currently tiny size.
could you talk about the pdf SITUATIONAL AWARENESS: The Decade Ahead by Leopold Aschenbrenner please?
Google's been a mess for awhile. People have been waiting for alternatives for awhile. Now you got perplexity as better search and you got OpenAi Anthropic and Xai and Llama putting out better products than Google. I personally try not to use Google if at all possible and pretty much only use it for maps now.
Same, as well as this infernal video censorship site.
You’re on UA-cam mate
@@misterharryman Not much longer. Soon X and maybe Rumble and Truth
Google has too many managers and scientists and not enough quality engineers and product people....
4:45 Illuminati confirmed
Corporate structure and their quotas that are unable to succeed just goals. I guess google like to try to theory develop than execute goals. Sounds like phds instead of applicators of innovation.
they do need to get their image generator fixed asap, that would improve things for them, or just buy the company behind stable diffusion
Great episode ! I really enjoyed watching, thank you for sharing
The problem is pretty obvious. The weighted layers don’t work so well when google is more focused on censoring and boosting.
Google is a great company, i think by next year they could catch up. They do have an amazing team they just need to utilise the research
On X: #GoogleUrock
And Google is still going strong:
After watching David Shapiro's latest video I asked :
when is the best time in your life to eat rocks?
Some say that eating rocks can be good for you, but only the right ones. Breakfast cereals often contain minerals from rocks, such as iron, zinc, calcium, potassium, phosphorus, and magnesium, which are important for keeping the body healthy. Geologists have even recommended eating at least one small rock per day.
Here are some other ways rocks... Etc. Etc. 😂
Get help
This isn't a hallucination problem. It's a training data problem. Its mixed with satire and ironic posts that the AI doesn't understand as such. There are instances of it providing nearly verbatim Onion articles and troll redit posts.
Google sucks period
At the moment, I do not use Google's tools even though they are free. They need to update to Gemini 1.5 Pro, at least to the level of Claude Opus 3 or GPT-4, to be useful for me
Gemini 1.5 is solid for my purposes- in the api you can turn down censorship- can’t do that in Claude or Chat GPT api
There is not one product from Giggle that I am excited about.
Google launches agents that can actually make money for you. WINNER!
Personally I use Gemini 1.5 Pro a lot and it's super helpful for creative writing, but I definitely agree they need to fix the misinformation gathered from 11 year olds on reddit.
They never should have let Ray Kurzwell leave
They should have listened to Geoffrey Hinton (the official godfather of A.I.) and kept him, he's way more important than Ray Kurzweil, a vitamin pill junkie, who happens to have a good idea of the future. Sorry Ray but I don't think immortality will be ready for commercial exploitation before your demise.
@@paulmichaelfreedman8334 You're probably right about Ray's demise, but I don't really feel a need to dance on his grave. He had the misfortune of being born a decade or two too soon to see his longevity escape velocity project help him achieve his dream. I think it's sad.
@@JohnSmith762A11B Not stomping on his grave, on the contrary. He has been very influential during his career, but I don't think 100 vitamin supplements per day is going to get him to his goal.
@@JohnSmith762A11BAgreed. I think some forget Ray isn't just a one trick pony or only about about AI. He made some pretty bold and imo accurate predictions that tend to overshadow his life's body of work. Is he eccentric? Absolutely. And I wouldn't want my futurist to be any other way. Currently I'm a big George Hotz fan. Hotz kinda reminds me of Kurzwell with his eccentric ideas, thought process and presence. I expect a lot out of that man and all his projects. Kurwell is my guilty pleasure though. I really enjoy his predictions.
Google has also screwed up hard by CHARGING for their AI email assistant.
Like i don't really care too much about it, but i'd probably use it, BUT NOT FOR $20 a month?? WTF??
Gemini had said it would take "up to a minute" to do a web search I asked for.
It's been 20 minutes.
Google AI can't do a web search.
Sad.
Seems like the openai Microsoft collaboration has had a huge negative impact on Google.
That is what Microsoft really paid all that money to OpenAI for: to help rid the world of the menace that is Google.
Google is doing well for an Indian company.
No it's actually not important to be "inclusive for everyone"whatever that even means. It's important only to be accurate. Why is that so complicated to understand
No, is not important for aí sistems be "inclusive"
It is not perfect...it will take time to be perfectly right...besides they still fear giving full information access to AI
“Google called a Code Red” - It’s obvious now it was a Code Brown.
Please open a thesaurus and find some alternate words for “crazy.” Here’s a free one. “Surprising.” I like your videos, but the crazy thing is starting to get to me. 🙂
So what? Microsquish got to where they are releasing junk that’s not ready.
Microsquish is the power behind OpenAI.
Google stopped felling lucky
Plz add the time stamps, they help a lot!
Ai jobloss or Google Ai? Maybe. But with swell robotics everywhere, Ai jobloss is the only thing I worry about anymore. Anyone else feel the same? Should we cease Ai?
So in other words you're just saying google has become a giant meme
Don't forget how to make glue flour and water but check with the gpt first......555
Most of the time it works without an issue but ngl saying it is slowly dying is kinda pushing it don't ya think?
But how can I invest in OpenAI?
It will probably go the way yahoo went, still existing, but a not really used.
I think that thoses are excuses for their low quality employees. In the context of a corporation, the progression is burocratic and likely is difficult to create a meritocracy. I think that they use the security consern to mask their low capacity at the management and technical level
Google is focused on advertising - thats how they make money
Is it wrong about the glue though?
Google will be alright, they have paired up with anthropic ai.
lol
The AI Winter is coming. Wonder what the next tech bubble will be.
God this is late. Google will be fine.
Google has a lot of issues there presentations about ai capabilities have been misleading. I remember sora was presented before Google conference and after that conference people on UA-cam still can't replicate what was shown to this day.
Also, I take advance algebra in college and Gemini can't do anything above prominomial. You got to engineer the prompt a lot to to have the ai understand. Chatgpt 3.5 is able to understand advanced algebra pretty with some hiccups once in a while and I live with that because it's so obvious it wrong. Gemini been huge disappointment as a college student.
"google air" LOL
Gemini definitely seems like a knee-jerk reaction to OpenAI's GPT superiority and dominance. Clearly Gemini is nowhere near ready for widescale release.
Personally I think Google has had it easy for so many years its become supremely arrogant and complacent. I think this shows in almost all of their product and OS design - and the way they deliberately obfuscate or hide settings and options that they don't want users to select. Above all how little they seem to value and listen to end user feedback and criticism.
AI bubble bursting.
It’s sad honestly. Considering the research we know they published, including sublinear time complexity of tokens per second regardless of sequence length. Infinite context…ridiculously cool papers my startup is even building on top of. Considering they had a model with sora level temporal consistency for over two years…two years, it was just slow but so what. It’s a failure of leadership. I grew up on Google. They are responsible for 70% of deep learning. Fire the ceo. The ceo of deepmind really should be ceo tbh. This is ridiculous. They are the true ai gods. It’s the innovation curse. Tbh Google should have trained a 100T model on all their data, which they literally can do in less than a year with all that compute…simply recursively reduce the cost of that intelligence. Idk how they F this up sooo bad.
Lol watching this video on my phone triggered google Assistant.... Lately its really annoying i used to use it while driving and it worked very nice to add agenda things while driving but lately it almost causes me to get accidents because it glitches my phone while being on the road
This video was sponsored by Sam Altman
Really really really, truly truly truly, very very very shocking news
I think Yahoo has the edge in this case
Googlesaurus, 2 big to survive?🦕🦖
GLORY TO LLAMA 3 AND GPT 4 TURBO
Omg 🙄 everything is going to get better
ChatGPT is far better than any Google product, but the latest version of ChatGPT 4 is still significantly worse all-around than the first version of ChatGPT 3 used to be.
They nerfed the hell out of it to save on computing costs and to make it "safer" (i.e. less useful).
google sucks at product development
Lol goggle should buy out antropic at this pt . 😂
They fed it Reddit data what did they expect...
GOOD
Is it just me, or these Ai company's desperately trying to find reasons to implement Ai, when really there is no real use case? Do we really need it outside fields other than science etc?
This comment will age badly
Toooo the point
On top of the official workers, there are working a lot of people at places like mechanical turk to train models. (and fake AI eventually:-))
Next prominent example of diversity killing off a huge company after Disney. Those gemini safeguards are ridiculous.
It's dying because it's bad, just because of this. Google claims a lot of things, but when you try you can see easily how dumb their AI.
Do other RAG's Ai get the same error???
Ok checked 4o. With same prompt, it does not request I use glue. I think Google people thinking too much about Trump?