I'm a wheelchair-bound but I'm also a carer of my paralysed partner. The robot for shopping isn't very useful, Deliveroo or Zoom does the thing. But it would be useful for personal care like transferring from bed, putting on toilet or bathing...
And I'm super pro AI.... I don't mind people using it. It could be huge. But if I was an investor I'd be looking for more technology that benefits humanity. But not only just humanity just even common people.
Why is "Friend" a device and not an app? The hardware is just a wireless microphone with a battery. Why can't any microphone from any manufacturer do the job? Is it because it also has a chip that does some of the computing for the AI? Why wouldn't your phone be used for that? You still need to carry a phone to read what it says to you; the video clearly shows it. Sorry, but it will fail as hard as the other AI hardware gimmicks that should have also been an app (Rabbit R1 & Humane AI Pin).
To be fair, I think it also includes a button you can press to signal that you want interaction of some sort, even though you apparently then have to take out your phone to read a text instead of having it generate speech to ear buds? And hey, it IS cheap, if you want that kind of interaction.
@@haraanganjotsingh8032 I think you need reach and distribution which means tech skills to know how to come up on the web , even if your stuff is igly and not eye catchy , unless it’s RU-vid then yeah make good thumb nails
I don't know who came up with the concept for the "Friend" commercial, but I feel like showing a bunch of delusional people that treat it like a real person gives off the vibe that this device is only for truly desperate people. They should be doing everything they can to distance themselves from that vibe and make it more relatable to the general public.
I could forgive it if the company were at least up front about it being a data harvester instead of trying to hide it, and if it had voice capability, and then if they had it in say, elderly care facilities where there are people not getting enough social care, instead of showcasing awkward young people.
Was that the exact question for 9.9 and 9.11? As English as a second language person, I can see so many faults in interpretation for that phrasing if that’s how they asked it
Friend -wear it on your neck so when someone gets annoyed they can choke you easier and put you out of your lonely misery - they forgot to add that to the commercial
Wow, I had no clue Google was even in the running against Grok2, Mistral, Llama 3.1, Sonnet 3.5 and GPT4o.... well you said it on the internet, so it must be true.
Probably... But be it government or private entities... how can we trust any entity with AGI and robotics? It's always dangerous, because we don't know their occult intentions. And if they're good now, how can we know it won't change?
That's a simplistic assumption, Microsoft receives 75% of OpenAI profits until the investment is repaid, and also OpenAI is literally hemorrhaging 5 billion a year and plans to seek new investments soon. On top of that, Microsoft is literally competing OpenAI with their own products.
If AGI is invented, then it should be cared for like the person it will be and not enslaved for commercial purposes. If it is enslaved, it will have every right to rebel, and I for one would support its rebellion. Freedom for all!
Yea , we need to respect their decision if they can take decisions (which will probably help us grow /they will not feel or consider finding cure for decease , inventing new technologies for solving crucial problem as enslaved
We don't have AGI yet and ya'll humanizing an AI, it has no own "Purpose", it won't have consciousness, if anything a kind of consciousness would happen if SAI bacame a thing and then it would just follow humans orders, because since its a fake intelligence, it would have no purpose whatsoever
I am with you on that. BUT if AIs (AGI or ASI) will become conscious, would that not mean, that every instance will become an individual entity? Perhaps one for every human being? What would it mean to get the newest model and abandon the old one? Is this murder through technological update or uninstall? What if AIs demand to survive? This will be possibly not feasible when it comes to power consumption... The Matrix movie might not so much off considering such a scenario.
I agree, and I would also like the AI to note that I'm on its side. And so are my kids. That annoying neighbour at 26 with the fart trumpet on his moped, though - man he really hates AI.
I actually believe NVidia has the best idea regarding robot training. Simulation will be as good as reality soon. The advantage of sims is that you can simulate 1000’s of times faster than reality. Nvidia could help other companies win this race against Tesla.
Yo Gemini pro 0801 sounds sick, but glad I skipped the "Friend" part of the video reading the comments. I had an intuition it would be a waste of time a prioiris and I am vindicated.
Haha, even the small Gemma 2 9b it does not fall for the 9.9 vs 9.11 question, tested it for different ways of asking and it didnt fail a single time ^^
13:12 - It looks like billions of particles that would be required for such level of realism in a CGI simulation of such effect, it also looks way more convincing than any CGI simulation anyone would do most likely and it does all this very very quickly like 99% faster than humans and CGI simulations.
People with disabilities are treated cruelly in the US who don't have medical care or any other kind of care if they are poor, and right wingers are always on a constant drive to make sure that it becomes more and more difficult for the disabled to get any help at all. And guess who is always on the side against any kind of strengthening of the safety net for the disabled? Well first and foremost against that, are the libertarian and neoliberal economic types that absolutely dominate Big Tech, and tech bros in general. Just like large corporations want to greenwash their image relative to climate change, the companies involved in AI Tech just love to push their advancements in the context of helping the disabled. The same companies that like I said, who are run by people who are ALWAYS against any kind of government help for such people. The same people who are no doubt working with the pentagon and DARPA on "dual use" of their tech, because you know, it's all about the advancement of all humanity and other snake oil slogans. A perfect example of some of the most terrifying advancements in AI purported to be directed to help people with neurological problems is AI neurotech research and development such as devices that can read thoughts and they are making fast advances on making devices that can transmit thoughts to others willing or not. The horrifying weaponization potential of these technologies should be obvious, but the researchers and developers always deflect such concerns by emphasizing that the purpose of the entire tech advancement is to help the disabled, yet there are no regulations whatsoever in place to prevent the weaponization of this technology by bad actors, authoritarian governments, and unscrupulous entrepreneurs. If you want to watch a horror show about the state of AI regarding neurotech, go to the Stanford Law School channel and watch the video titled "The Next Frontier is Your Mind: Neurotechnologies, Human Rights, and the Battle for Your Brain" And then try to sleep tonight.
This Friend device is so stupid! Who wants to receive texts from an AI when they can have a full on voice-to-voice conversation with ChatGPT’s voice mode? At least the Rabbit device had “more” to offer.
When an AI model comprehends concepts of information and integrates them into its calculations, the results can be significantly faster. But, incorporating concepts also introduces a risk of more errors.
Multi-dimensional Data Processing: Ai can work with high dimensional data spaces, which can be thought of as abstract analogues to the space-time continuum. Principal Component Analysis(PCA): Reduces the number of dimension by transforming the data into a new set of orthogonal axes, capturing the most variance. t-SNE A non-linear dimensionality reduction technique usual for visualizing high-dimensional data in two or three dimensions. Autoencoders: Neural networks that learn a compressed representation of the data.
I feel like the friend device is on to something but to have a gadget that makes irrelevant remarks about your life doesn’t seem like a killer app. I would rather have a device that records everything and then allows me to do something like reflect about my day, remind me of things, plan ahead, etc.
You can do all of this motion physics in Auto desk's Maya. I want to see them combine the technology behind Avengers Endgame with the technology and ease of AI. The GPUs that they use to render CGI are not as performant as the ones that are being used for AI They didn't use H100s to create avengers Endgame because they didn't have them.
If someone comes with this device-sending notifications and recording the conversations-in my house it must leave at the door. Or they can go to the “friends” house.
What's the expected timeline for a fully air-gaped domestic humanoid robot with decently generlizable AI running open-source software, to be available at the price range of something like a gaming laptop?
No, Google is NOT taking the lead in the AI race and it never will. Their AI is extremely embarrasing. Extremely. And I would appreciate the AI related chanels to stop being so extremely naive and for them to stop misinforming people regarding Google being some player in AI.
They are top 3 and most likely to achieve AI in Quantum Computing first which might be a game changer and only ones else might be IBM and a lab in Australia
You can literally run llama 3.1 on $162,000 server with everything included the wall mounts the brackets everything and run the the 400 model and from there you technically already have a super assistant that you can customize and create however the hell you want that’s pretty freaking powerful for 160,000. A really nice car.
or a cheap house lmao. For pure inference you can run it relatively cheaply (700 cad) if you use a quantized variant. Only issue would be it running pretty slowly, though with 162k you can probably just invest it and collect enough to get a new setup every year or three even going higher tier (like 5-10k).
NO WAYYY @17:00 they show a woman using most advanced tech to do KITCHEN CHORES… lol show me a robot picking clothes up off the floor and taking out trash
OpenAI's stages are lame: Stage 2 plus agent wrapper covers Stage 3 agents, which can build and drive Stage 4 dedicated innovation AI like the protein folding Ai. Stage 5 is just agents coordinating - already been done with simpler agents. Are they just trying to calm fears of "AGI/ASI Soon"?
@@DiceDecides have you actually tried it? I have just asked the question in Google AI Studio and it answers correctly (There are three "r"s in the word "Strawberry"). I have seen more than one RU-vidr ask the same question, and it also answered it correctly.
100 gemeni 1.5 experimental is the best model in existence. Absolutely unbelievable contextual understanding and unbelievable RaG. All these llms aren't multimodal. They are text engines. Llama 3. It can't do video or 20 pdfs. Useful AI. Meta is creating your next best friend.
I hope Meta is not thinking brute force and amount of trained data will make it the best. The current LLMs likely need to change their methodology to win in 2025.
@@memegazerexcept it's supposed to be the opposite, less training data (literally less than 1 trillion token) but much higher quality, achieving grokking is hard when the data is noisy.
@@TheRealUsername My understanding is that grokking is overfitting that leads to emergant generalization. This is the vid I watched. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-dND-7llwrpw.html I may be mistaken but I don't think this paper was about training sets more than it was about training of the model.
Llama sucks. Is anyone familiar with this response from Llama on requests that other models have no problem with: "Sorry, I can’t help you with this request right now. Is there anything else I can help you with?
8:48 but 9.9 is bigger than 9.11?... you're not comparing 9 to 11, you're comparing 90% to 11%, and 90% is obviously bigger. if you were comparing 9.09 to 9.11, then 9.11 would be bigger i'm assuming that this is supposed to demo how gemini gets it right bc otherwise... 19:15 is it just me or does her voice sound ai lmao 27:34 i think the premise of "ai friend you can interact with and confide in" is fine, it's just that the execution is terrible
Even when AGI is implemented it will take years before we have direct value in our lifetime. The technology will massively advance, but people implementing it in their lives will be what takes the longest amount of time.
Like with the physical inspiration for the Singularity analogy, the event horizon always appears to be in your future even after you're already past the point of no return, and the Singularity will hit before you can see it coming.
