This promise of a world responsibly empowered by AI continues to drive our work at Google DeepMind. For a long time, we’ve wanted to build a new generation of AI models, inspired by the way people understand and interact with the world. AI that feels less like a smart piece of software and more like something useful and intuitive — an expert helper or assistant.
Today, we’re a step closer to this vision as we introduce Gemini, the most capable and general model we’ve ever built.
Gemini is the result of large-scale collaborative efforts by teams across Google, including our colleagues at Google Research. It was built from the ground up to be multimodal, which means it can generalize and seamlessly understand, operate across and combine different types of information including text, code, audio, image and video.
↫ Demis Hassabis on Google’s official blog
It’s no secret I’m not particularly impressed by “AI”, not least because its ability to autocomplete complete nonsense based on copyrighted works it’s drawing from without permission and the dangers this might represent to our society. That being said, Google’s new “AI” thing, as demonstrated in this video, actually seems a tiny bit impressive. It still looks like to me like it’s just blurting out random information using fairly mundane things like object and speech recognition, but the fluidity of it all definitely feels a lot more natural than whatever OpenAI and Microsoft have shown so far.
I’m still not even remotely interested in any of this stuff, but this at least seems slightly more possibly useful than other examples I’ve seen so far.
And what great works or world problems are Google deploying their amazing AI to solve?
Ah yes, they pointed it at their biggest commercial opposition in the space (openAI) to discredit them..
Thom Holwerda,
You’ve been very critical of AI in general, I sense some bias but I don’t know if this is fair to say. Ultimately though the computers are clearly getting better at “human problems” and we’re reaching the point where humans are less special at human skills than we used to be. Whether AI is capable of displacing humans is becoming less a question of skill, and more one of ethics and cost.
@Thom Holwerda & @Alfman: I love both your insights. I’m also rather sceptical about AI (I believe the AI-advocates have used the definition of IA for AI to prove it works), but then again, I’m also very influenced by the paper “No silver bullet” (Fred Brooks) -> http://www.cs.unc.edu/techreports/86-020.pdf (read 4.3 and 4.4, and you’ll probably want to read on :-)).
On the other hand, here in Belgium, there recently was a study that said that for the first time the next generation will be dumber than its predecessor, and that might actually bring Alfman’s truth closer to reality.
klodder,
I’m not familiar, but I’ll try and read it some time.
Here in the US I think that may have already happened, we’re just ahead of the game 🙁
https://www.msn.com/en-us/news/world/us-10th-graders-score-lowest-ever-on-international-math-test-whole-world-is-struggling/ar-AA1l37Tn
Not only are we bad in math, but we’re weak on basic facts and history too. Trends like this are why I’m always wary of comments that downplay the Chinese as a bunch of non-innovative copycats. Western companies and influence remains stronger for now, and so they copy us, but how long do we have before Asian countries with the educational advantage take the lead and we have to copy them? I don’t think it will be too long before they’re out innovating us ether this generation or next. Even with good policies, it takes decades to change this momentum and we’re headed in the wrong way. Oh well, this is all off topic.
Yeah, the demo is impressive. But it also kind of reminds me of the Siri promotional ads Apple put out. https://youtu.be/nqXGWQhowXk?si=5mhFFmvPhj9GAsOH
It seemed at that point very impressive and I thought AI assistants were a few years away. The real life experience of using Siri was very different with the iphone 4. Even today using it can be very frustrating for all but the most basic tasks. Is the google ad closer to reality than that siri advertisement was? I don’t know but I suspect its not. Working with AI right now is a bit frustrating a lot of adding more and more specific prompts and tweaking of previous prompts until you either get the result you need, it starts making up bullshit, or you get frustrated.
Bill Shooter of Bul,
That was more akin to a PDA with voice assist than AI though.
Perhaps it’s getting too good at mimicking humans then? Haha.
Don’t get me wrong, you do raise good points, but we would still find faults even if it cloned human behavior perfectly. So the question might be whether we should be using human standards to judge AI? Is that good enough?
This level of LLM AI does basically parrot back what humans have created. mixing it in a way it things most people will like. But all of that is subject to interpretation. I don’t want an person of average intelligence answering my calculus questions.
Bill Shooter of Bul,
We already have math programs that can do that quite well. The point of LLM generally isn’t computation but interacting in a human way.
This is what I’m basically talking about in the review
https://techcrunch.com/2023/12/07/early-impressions-of-googles-gemini-arent-great/
Of course the demo google selected to share looked impressive. The reality is much less so.
Bill Shooter of Bul,
The technology is still young, and the cutting edge naturally has bugs to iron out. Still, a couple of hiccups here and there doesn’t mean that AI stops getting better. We shouldn’t underestimate AI’s potential, it will mature.
Good to see Google shipping a press release