When it happens, I think we'll get used to it pretty fast and we'll be like, "Oh, that's all it is."<p>Then everyone can order groceries and plan trips by voice conversation with a smart assistant that actually knows us.<p>And it will become productized and have tiers and specialities but AGI is not going to be any more "OMG my life is changed foreveaah" type watershed than, say, the iPhone.<p>Don't. Believe. The. Hype.<p>That's my take. What do you thinK?
I believe A(G)I will be like a really dumb human with about 10 thousand years of experience in a certain specialty. It's hard to predict how things would change, but imagine security guards who have memorized every face in the country.<p>However, don't overlook that the iPhone did change things forever. The difference in lifestyle between 2020 and 2000 is more than 1950 to 1990. Your life might not be different, but it makes a difference to the factory worker who had no savings, but is now an Uber driver who can freely work 16 hours one day and sleep the next day.<p>We'll likely move a "class" upwards, in that humans will no longer be doing brute labors and driving trucks, but still be scheduling truck routes and telling the AI what to do. Humans will probably be more involved in educating and training, maybe even disciplining stray AI.<p>I also don't believe we can achieve proper AGI without inventing some kind of emotional state. Maybe AI will develop crushes on their trainer, not so much for sex, but out of gratitude and for survival. Or there could be anger, ambition, envy, which might simply stem from the desire to learn and experiment. Maybe AI may become sweet and manipulative as it predicts patterns in how humans respond. There will also likely be some equivalent of dopamine, and AI could be addicted to staying above 80% charge and so on.
I’d say we are at the beginning of a lot of technological trends, trends that will span centuries, and we don’t quite understand how they will play out.<p>AGI probably won’t get <i>really</i> interesting until 50-100-150 years from now. Ditto for stuff like space travel, self-driving cars or social media.<p>I’d also question your assertion that the iPhone wasn’t a revolutionary change; we simply haven’t understood the full implications of mobile computing and mobile access to a global network. Remember that the Internet has only been around in a consumer form for roughly 20-25 years. This is an incredibly short period of time, compared to say, the printing press or the steam engine, and we’ve already seen society completely transformed.
The first thing that comes to mind is a question. If we made something as smart as you why would you be legally allowed to use it as a slave?<p>Next the logical argument is that technology doesn't stand still. If we start with a human level AI we can gain increased runtime simply by throwing more compute at the problem. There is no particular reason we have to run it in real time why not 10x or 100x or 1000x.<p>What would you produce in your field if we gave you a century or a millennium between now and next year?<p>This leaves the last and perhaps obvious point that an AI that can improve itself can use these centuries of run time to work on such improvements.<p>If you have human level AI tomorrow you don't have merely human level AI for long and once something surpasses you, you are poorly positioned to predict or control it. This is not to say it would inherently be malicious but does your pet cat get much of a vote in the running of your day to day life?
I wonder if gpt-3 is our very first instance of AGI and it’s obviously very bad and barely passable. But it meets some definitions. Can learn some new tasks without retraining. Applies common sense to some questions. And can do some arithmetic.
<wild speculation><p>AGI is not the same as human intelligence. It has the generality of human intelligence but since it isn't restricted by biology it can scale up much easier and can achieve superhuman performance in pretty much any individual task, group of tasks, or entire scientific or technological fields. That's pretty exciting.<p></wild speculation><p><reality><p>It's questionable whether the above is possible at all. In all likelihood none of us will see anything even remotely close to this in our lifetimes. We're currently so far away from it that we don't even know how to get started on solving such a problem. Nobody is currently working on this, despite how they're advertising their work.<p></reality><p>I guess what I'm saying isn't that AGI will be underwhelming, it's that it won't exist at all, at least as far as we are concerned.