Google, Other OpenAI Rivals Make Their Own Big Announcements (tomsguide.com) 19
Thursday OpenAI released a "smarter, faster" ChatGPT. But there's still competition, notes the tech site Tom's Guide (which is liveblogging December's AI news). "Not to be outdone by OpenAI, this week has seen several big announcements by other AI companies."
Google Deepmind unveiled Genie 2, a tool capable of creating limitless 3D environments. It could create playable games based on a single text input.
ElevenLabs announced a new Conversational AI system. It's a voice bot meant to feel like you're making a phone call. Tom's Guide AI editor Ryan Morrison used it to clone his voice to act as technical support for his dad.
OpenAI will probably announce an upgraded Sora video model in the coming days, but we were impressed by the new Hunyuan Video model that released a demo this week. Sora has some serious competition and we're interested in seeing how it competes.
ElevenLabs announced a new Conversational AI system. It's a voice bot meant to feel like you're making a phone call. Tom's Guide AI editor Ryan Morrison used it to clone his voice to act as technical support for his dad.
OpenAI will probably announce an upgraded Sora video model in the coming days, but we were impressed by the new Hunyuan Video model that released a demo this week. Sora has some serious competition and we're interested in seeing how it competes.
another day another voice bot for scammers (Score:2)
Re: (Score:2)
Or the bot could be the next scammer. Deepscammer.
Re: (Score:2)
I'm looking forward to the AI developing its own scams, which it will try to pull on the AIs we hire to screen all communication because of the flood of AI generated crap we have to wade through to get to any real communications from actual humans. What a glorious fucking future we're building, where most resources will be poured into generating crapfloods of non-information so that a few companies can keep the profit churning.
Re: another day another voice bot for scammers (Score:2)
Umm, Dear {ai of your choice}, how can I convince grandma to wire me money?
I've always said the primary use case for AI is fraud.
Fakery and hallucination at lightning speed (Score:2)
I can totally see the appeal for low-effort professionals and grifters.
Re: (Score:2)
Joke of the day.
Re: (Score:2)
"I don't want AI that rewrites history, telling us Vikings were black, that Henry 8th was black."
Neither vikings nor Henry VIII are examples of good people, so why would the left wing want to portray them as black?
Re: (Score:2)
Big Hairy Announcement ! (Score:2)
Guaranteed . Will be berry exbensive too. Don't use 3 month old AI. You want new AI. We got new AI. Subscribe now.
Also, coming soon : more announcements for more new expensive AI.
Did I mention it's really really new? And that it's AI?
Are they getting any smarter? (Score:2)
Are they getting any smarter? (Warning: chemistry nerd post.)
Look, I love to talk shit about chatbots, and they have been slowly getting better. But last night I asked ChatGPT-4o about how to make dish soap given a list of ten available detergents and an alkali. I'm a dilettante cosmetic chemist, but its answer was laughably bad--not that I expected a worthy successor to Dawn, but some of its mistakes betray the fact that it doesn't have a real knowledge model. For example it said 1% baking soda would add a
Re: (Score:2)
It's just putting words together. Of course that's what humans do,
Most humans have the idea first, then push that idea into the form of words. ChatGPT only has the form of words. Kind of amazing what it can do with that.
Re: (Score:2)
Right, but what I meant is that we have decades of inputs and a truly unfathomable amount of internal feedback (creating concepts/abstractions, biases, reinforcing, weakening or forgetting) before ideas are generated. Chatbots get the same amount of input or more, but their processing is weaker, or they have an architecture that's inefficient at processing of this input. Some of the rules of thumb that are easily accessible, for example all the reasons that baking soda would never be an abrasive in dish soa
Re: (Score:2)
Re: (Score:2)
Is your point that chatbots have more input than I've had? I know. But they don't store (or continually reference) every input, so the proportion of crap to relevant data matters. It gets lossily compressed, in a manner of speaking. My brain also lossily compresses everything, but the specifics are quite different.
Re: (Score:2)
Re: (Score:2)
Ahh, okay. We are beyond my ability to talk about chatbots with any confidence. I though the data manipulations (transformations between vector spaces, matrix multiplications, etc.) that the chatbots do could be considered analogous to thinking, but I don't know enough to argue that convincingly.
Forgive me if my "faith" in this is offensive, but an important idea that bears mentioning is that the conscious experience of a thought is not the information processing itself. We will some day be able to function
Re: (Score:2)
I think you may be thinking the experience of a thought is part of the pipeline that produces actions, like "sensory input -> neurons -> ??? -> conscious experience -> ??? -> neurons -> actions like speech".
Nah, I can't get that detailed. I assume that it's all neurons, including the speech. I just say there's some way in the brain where ideas are stored separate from language. In particular, you can feel this when you learn a second language, and you come up with the idea, and then there is a brief moment when you have to decide which language to use to transmit the idea.
idea in neurons -> words produced by neurons
And for the computer, it's more like:
statistical evaluation of most common next word -
Re: (Score:2)
The mistake most people make is thinking that chatbots operate on facts and concepts. That's just part of the trick. They're designed to give output that gives that impression. In reality, they operate on statistical relationships between tokens. Once you understand what that means, the often laughably bad output makes perfect sense. Those aren't 'mistakes' or 'hallucinations', they're exactly the kind of thing you'd expect.
You know a lot about chemistry, so you're quick to notice when it produces nonse