The difference is Disney doesn't have any factories. I'd bet good money Tesla has Optimus doing real work in their factories, at scale, long before anyone else.
And it's one thing to have a cute demo showcased under ideal scenarios and another to have it deployed in mission critical environments around real people.
We have had robots stamping parts for years. There is no need for a humanoid style configuration. Meanwhile what is needed and far more complex are robots that can interact with unpredictable people in unpredictable environments.
Boston Dynamics has already demonstrated they can do this.
You would just change the machine that the human uses. If a task is too difficult and hence needs a human yesterday, and if there is a solution that a humanoid robot can now do today then you would simply get rid of all of the humanoid robotics surrounding the core problem. Even if that means bolting the humanoid to the floor just to use that single part you need.
Elon is the guy who said, "that is the craziest idea I've ever heard, here's billions, let's build it"
It's 100% Musk that this happened, literally no one else would have done it. Everyone thought it was dumb until today. No company, no government or sci fi movie for that matter had even thought of it in the first place.
SpaceX engineers given the green light by Musk, not NASA, made it a reality. That's not to say Elon didn't make many decisions along the way in the design and development of both Starship, the launch towers and everything else.
The point is, Elon deserves a hell of a lot of credit along with everyone else. Everyone has their part to play in the success of the mission.
You’re pretending if you think anyone but Elon would approve building a massive launch tower with chopstick arms to catch what is essentially an incoming bomb.
Because it's Elon who said "ok, build it." There's no one else except him with that power and/or the guts. Even landing a rocket the government had given up on, and no company was even trying until he did. People were calling the 'chopstick' landing system a dumb idea until today.
It's not like engineers don't come up with wild ideas all the time to their leadership, but is the leadership good enough to understand the good ideas from bad ones? Take the risk, spend billions to actually execute?
Elon has enough of a physics/engineering background to ask the right questions, understand the trades engineers put in front of him, and make the risk/reward calculation to make the right decisions the ends up winning.
To get what SpaceX has you need strong technical leadership all the way up the chain. Many companies don't. Their CEOs are experts in legal, PR, finance, etc... They make poor technical decisions.
It'd be interesting if games can be revolutionized with realistic worlds and NPCs with the use of a dedicated NPU. Like how game graphics were revolutionized with the GPU.
Maybe, but I think that more handcrafted and tailored NPC interactions would be much more engaging than AI rng (not to mention way more energy efficient)
Even oblivion had pretty lively NPCs, if quirky.
Majora’s mask takes it to the extreme by scripting 3 in game days for each and every character.
I think AI generated voice "acting" will be great. Someone made an addon for world of Warcraft maybe a year ago that gives NPCs voices when giving you a quest. While not all of the voices exactly match the character, the voices themselves seem pretty incredible.
Looks like even for the non-realtime API they're charging $200/M for output audio. Their current TTS API is $15/M (characters) for output audio, which equates to $60/M if each token is around 4 characters. Then add in the manual piping to the 4o LLM which is $15/M, around $75/M total.
So from $75 to $200/M is a big premium for the convenience of one model and the quality of multi modal input/output. Will have to test and see if it's worth it.
Also is there still no way to connect users directly to OpenAI? Like directly from a user's browser to OpenAI's servers, without the user having to supply their own API key? How does this work with realtime that needs websockets? Do I need an intermediate proxy server for all my users conversations? Seems like a waste of bandwidth, an unnecessary failure point, and a privacy problem. I hope I am wrong.
To stick with the metaphor, it's not super unclear if agency is within grasp or on the other side of an abyss. LLMs are definitely an improvement, but it's not at all clear if they can scale to human-level agency. If they reach that, it's even more unclear if they could ever reach superhuman levels given that all their training data is human-level.
And finally, we can see from normal human society that it is hardly ever the smartest humans who achieve the most or rise to the highest levels of power. There is no reason to believe that an AI with agency would an inherent "it's over" scenario
What is happening right now is so obvious that people have been predicting it for over 60 years. It is ingrained in our culture. Everyone knows what happens when the AI becomes smarter than us.
If you 'see no reason' then you are looking through a microscope. Lift your head up and look around. Agency isn't black and white, it is a gradient. We already have agency to some degree, and it is improving fast.
We kill millions of humans with agency every day on this planet. And none of them immediately die if we stop providing them with electrical power ... well I guess a small amount of them do. Anyway, we'll be fine. If the AI come, can they immediately stop us from growing food? How can an AI prevent my orchard from giving apples next year? Sure, it can mess up Facebook, but at this point that'd be a benefit.
How can you shut down something that can multiply into data centers in different countries around the world?
I'm not sure you realize this, but computers control everything. A dumb bug shut down windows computers around the world a month ago. A smart AI could potentially rewrite every piece of software and lock us out of everything.
Those medicines your friends and family need to stay alive? Yea the factories that produce them only work if you do what the AI says.
I don't want to be a hater but according to the UN about 61 million humans die per year, so that only comes out to ~167k per day rather than millions. Most of those will die from old age too, rather than "being killed".
Your main point is true though, even superhuman AI would have a rough time in actual combat. It's just too dependent on electricity and datacenters with known locations to actually have a chance.
I’m sure the superintelligent AI will convince humans to transport its core while plugged into a potato battery. But honestly did the supply chain attack on Lebanese pagers last week teach you nothing? AIs should be great at that.
> We kill millions of humans with agency every day on this planet. And none of them immediately die if we stop providing them with electrical power ... well I guess a small amount of them do. Anyway, we'll be fine. If the AI come, can they immediately stop us from growing food? How can an AI prevent my orchard from giving apples next year? Sure, it can mess up Facebook, but at this point that'd be a benefit.
1. IMHO, genocidal apocalypse scenarios like you describe are the wrong way to think about the societal danger of AGI. I think a far more likely outcome, and nearly as disastrous, is AGI eliminating the dependency of capital on labor, leading to a global China Shock on steroids (e.g. unimaginable levels of inequality and wealth concentration, which no level of personal adaptability could overcome).
2. Even in your apocalypse scenario, I think you underestimate the damage that could be done. I don't have an orchard, so I know I'm fucked if urban life-support systems get messed up over large enough area that no aid is coming to my local area afterwards. And a genocidal AI that wasn't blindingly stupid would wait until it had control of enough robots to act in the real world, after which it could agent-orange your orchard.
1. A massive loss of jobs like that would lead to a better redistribution of wealth. I don't believe the US would adequately react, but I trust the rest of the world to protect people from the problem.
2. Which robots?? Until robots can, without any human input, make any sort of other robots, then there is nothing to fear from AGI. Right now they barely walk, and the really big boys can solder one joint real good. That's not enough.
> A massive loss of jobs like that would lead to a better redistribution of wealth...
How do you come to that conclusion?
I think we'll get massive inequality because it'll be far easier to develop new technology than change the core ideology of capitalist society through some secular process. If a "better redistribution of wealth" were in the cards, it probably should have happened already, and probably cannot happen after the 99% loose most of their economic power by being replaced by automation.
If the 1% are forward-thinking, they'll push a UBI scheme to keep the 99% from getting too disruptive before they become irrelevant.
Multimodal generally refers to _just_ two (or more) modes. Omni-modal was invented by OpenAI to describe their approach. I'm sure Anthropic is working on something similar to gpt-4o but I don't see how that invalidates my argument about anthropic doing well on benchmarks. In particular because their model was released several months _before_ gpt-4o and still holds its own fairly well.
Personally I like Anthropic, they do one thing well, but there are a lot more applications for AI that OpenAI is offering solutions for. Anthropic, not so much.
Given that ChatGPT is already smarter and faster than humans in many different metrics. Once the other metrics catch up with humans it will still be better than humans in the existing metrics. Therefore there will be no AGI, only ASI.
