Is good AI a hardware or software problem in our age?
> computers are so dense and small, working on a nanoscale
> extremely large amount of transistor or calculation units, complex computations in less than nanoseconds possible
it would seem that we have more than enough processing power to simulate good AI and just lack methods for learning/training it, however
> human brain seems to be dense, packed with many neurons
> human brain is nonlinear and mostly working in parallel
Is good AI a hardware or software problem in our age? > computers are so dense and small, working on a nanoscale
Is good AI a hardware or software problem in our age?
If the goal of AI is to replicate the human brain, then we need to understand how the brain works first. We don't have the foggiest idea how neurons encode and process information.
>If the goal of AI is to replicate the human brain,
I don't think so. AI aims to do tasks that are at least as complicated as the ones a human could solve.
I dpn't see why that would have to be the case any more than how well we needed to understand how horses run in order to make cars.
>We don't have the foggiest idea how neurons encode and process information
consciousness is not a material phenomenon and does not come from the brain
the brain is a satellite or receiver of signals
>We don't have the foggiest idea how neurons encode and process information.
I maybe we won't ever
I feel in the future we might use brain matter as computers but an artifical neural network would have to be attached to it to send and analyze data .
The secret is folds
It's both. I think cyborg computing is the future. That will actually take us into the next technology age.
I work at an AI startup and I’d say the main limitation to advancing AI faster is that most people hired to work in AI are dumb as shit and can’t design a simple ensemble, algorithm, analysis, etc to save their damn lives
Well, think about it, how much statistics is even taught in a standard undergrad program in even mathematics? Maybe a class. Not even getting into computational statistics which should be a pre req for pretty much every STEM degree (it's amazing that people can get any science or engineering degree and not have to learn applicable statistics).
If you don't understand even basic computational statistical techniques, you will never really understand AI.
statistics as a field is literally and unironically fake and that's why no one bothers teaching useless delusional tripe
if the conditions for a ball to land in a hoop are met, it will land in that hoop every time
Statistics is as fake as fitting a line to data. You don't know what you are talking about because you've had a piss poor education in statistics and probability. Don't worry, that's average for the modern STEM degree despite it being literally the most valuable skillset you can have in any of the fields.
let me guess
> everything's fity-fifty, probability does not exist
> everything that will happen is bound to happen deterministically, no chance involved
> no, I can't tell you the weather tomorrow for certain, your model is wrong actually every weather is possible tomorrow
The point of a university I'd just to prove thst you can learn bullshit esoteric topics, 99% of things learned even in a stem program never comes up again.
>I work at an AI startup
How do you get into that?
My job has had me do computer vision projects and that was interesting abd making me want to pivot into that.
>>I work at an AI startup
>How do you get into that?
>My job has had me do computer vision projects and that was interesting abd making me want to pivot into that.
That's a good background, I started in computer vision projects and I was doing something that combined computer vision and nlp and that got me in to nlp.
Can a computer have an internal vision realm, like a video game space, and be given the prompt to explore and play and create things, without nessecarilly having a goal?
Like a learned computer that has seen many images, games, principles, videos, formulas, etc.
Give them ms paint, or some Lego space realm with infinite Legos and tell it: play around, draw, build, whatever you want
Technically yes but practically no, because an AI without a goal will spend all its time doing useless things, and as soon as you give it hints about what things are more useful to do over others then you have biased the AI towards doing that and it’s no longer doing things at random without a goal because you’ve given it one.
My dad works at Nintendo and he says it's actually a hardware problem.
Hey how can a dumb as shit non-programmer/engineer/researcher get a job at an AI company?
Why would a non-programmer/engineer/researcher want a job at an AI company? Anyway you can probably get in just by lying about your programming skills on your resume and dropping a few buzzwords during your interview, because for some reason the people with money in AI have no desire to actually evaluate the skills of the people they hire. In the rare case they sit you down with a technical person who might be able to call out your bullshit, start dropping hints you may sue for discrimination
Lmao. Mainly because I find the space interesting but am a more of a product marketer by trade.
Oh I was under the impression you were a skill-less pajeet trying to enter AI for the money.
Let me tell you, there is a HUGE need for product marketers in AI. There's some good AI products out there and there's some bad AI products out there. Good AI products need product marketers to help clients understand their value. Bad AI products need product marketers EVEN MORE just to move the shit they had 10 pajeets waste their time on for 6 months. If you can market Good AI Products you're a good markerter, if you can market Bad AI Products then you are worth your weight in gold.
Oh shit really? Any advice on finding good AI companies/products? Either way that gives me way more hope than I had before of actually finding something in this space.
Honestly every AI company I know is just constantly looking to expand their marketing and sales departments. Where are you located? Do you have a portfolio? With a portfolio you can probably go for remote work. If you don't have experience then you probably need to start in person. If I were you I'd walk into any coworking space and just hang out near the water cooler and look for a C-suite type, they're usually the white or asian males who are over 40.
>Is good AI a hardware or software problem in our age?
Understanding and create ML systems is a problem that goes beyond just hardware and software. It's first a set of techniques and theory. Hardware and software were isolated concerns for ML in the past, but the systems CS people have made a strong case for hardware-software co-design and accelerator design.
A small but important example is leveraging GPU/CUDA for specific ML implementations.