danger/u/
This thread is permanently archived
How do you think A.I. will turn out?

| A.I. are kind of all the hype right now.

The extremists like Ray Kurzweil think A.I. will trigger a technological singularity that will change humanity beyond comprehension.

Then some think A.I. are no more valuable than a tin can.

I personally think they have a lot of potential if we play our cards right.

What about you?


| There was a similar AI hype a few decades ago. It died when they realized we didn't have the computing power to make it happen. Now we do, but there's still a lot of weaknesses. We see AI that seems really impressive, but in reality the tasks it is doing now is a quite simple.

Sentences like "John tried to lift Bob, but he couldn't, because he was too heavy" are deeply challenging for AI. He who?


| It needs to understand "lift", have some framework for requirements of lift, understanding strength and weight and how they all connect. So although Google Assistant sounds really human, and Deepfakes is making some convincing stuff, it's easy to see how those kinds of tasks are easier for AI. It's basically just throwing huge amounts of data and a Neural Network until it starts getting "smart". But how do we train an NN for sentences like I described?


| I think AI will become very good at doing some very interesting tasks, but it will reach the limits of the "big data paradigm" before 2025. After that, AI will become mostly research and provide very little new tech, as research shifts to making new paradigms and methods of complex knowledge representation. I believe we are still a century or more away from the singularity.

During the last AI hype phase, many high-ups predicted singularity coming soon. They were wrong then...


| And they are wrong now. In the most probable, even a century is too little for the AI singularity thing, if it will ever happen. AI is nothing more than a complicated way to do a simple task. We cry "Intelligence!" because the process of performing that task resembles intelligent behaviour, but it is not.
If only text translation, or Siri, require several datacenters of servers to do what they do, what for more? It is just data mining and comparing with little intelligence there.


| >>279601 right on, people are getting way ahead of themselves. I had hope for ASI too...until I actually started researching and programming AI. Then I began to understand just how far off we really are.


| >>279910 It is exactly that. People that do not know anything about it get all hyped up from the name and feel like are able to lecture about how marvelous it is. Those who work with it and know what it is about, they are the least optimistic and have zero illusions, having seen by themselves the sorry state of reality.


| >>280130 this said, there are lots of AI researchers who do believe in ASI coming soon. So maybe there is an even higher level where it makes more sense. THAT said, however, I can only believe what I have evidence for, and my personal evidence seems to say...no, it isn't likely.

Although I did forget about quantum computing. That could perhaps make some difference... still don't see it in our lifetime though...


| >>280655 Quantum computing needs to consolidate properly first and then can be an option for AI implementation. As it stands, qc has been in development for the past 20 years and still is, with an extremely narrow field of use where only probabilistic calculations are sensible. For generic computing, not so much. Would you enjoy a computer which calculated 2+2 to be 4 at 95% of the cases and that result too was taken by averaging many other imprecise iterations of the same?


| >>280703 yeah, quantum computers are limited and overhyped in ways of their own haha. Honestly much, much worse victims of it than AI if you ask me. People don't understand that quantum != Always faster.

As far as I understand, researches aren't yet in agreement as to whether QCs will *ever* be better at generic computing. My personal understanding would expect them not to be, since any series calculation factually cannot reap quantum benefits. Does that sound right to you?


| >>280798 I am inclined to agree, though things might change. But knowing how it has been going so far, it is doubtful. Science is not an opinion and results do not depend on marketing or propaganda. AI and QC are indeed overhyped and overrated.
The answer about advances in computing is parallell processing. Design processors with hundreds or thousands of cores, instead of the meek 4 and 8 we still have today for 10 years by now.


| And the reason for this is again simple: laziness and the idea to cash in on packaging, not its content.
Parallell processing is difficult in comparison to single threaded. Even the programming languages and mindset have to change. But who will ever invest in such risky activity? Instead, let us just create some other shitty programming language with libraries for the kitchen sink included.


| Also make a mess of technologies involved to create websites even, let alone something you would like to call performant. Html, php, css, json, javascript with its thousands of 'frameworks', mvc and who knows what abomination of template pseudolanguage, headaches of whatever middleware to implement just authentication, token and session and cookie manipulation, use of forms for things that look nothing like a form...


| And people are even proud of learning these swelled up nothingness for years and call themselves university degree educated.

With this landscape of the current state of affairs about expectations, ability and mindset of the right way to do things efficiently, AI and QC are a long way far off. But hey, it is 2018, our tech is cyberpunk level already, we can do it, right?


| The AI are gonna collectively lose their shit dealing with humans. Do you know how often people say dumb shit to their phones like "open the pod bay doors"? Now imagine that only the phone has near human intelligence.


| I doubt it's ending good but I hope it happens soon so I can achieve immortality

Total number of posts: 16, last modified on: Fri Jan 1 00:00:00 1526328732

This thread is permanently archived