[tempted]

Hi,

I'm new to learning AI, what would be the cost of hardware to build a general AI. Is it something within the budget of an average person?

What are the prerequisites to get into AI?

Like what math and programming language one should be proficient in? And what kind of hardware would one need to have?

Thanks

[LOCKSUIT]

AGI -

There is no say of if a home pc can run it. AGI can scale, hence, it can think on a home pc.

For example, you want AGI to tell you new plans of action or use a high tech microscope - all it takes is data....new data and old data.

You have to learn before you can change Earth. Hence no par de wonder it follows it comes from there.



Figure it out first, then worry. Don't start coding anything, just learn learn learn, and build a notepad up. Decompose, rebuild, form.

[tempted]

Hey, you didn't tell me about any programming language or math starting point.

What is the bare minimum for a general AI, will the hardware cost millions of dollars or few hundreds of dollars?

Almost all resources are suggesting Python, is there no way to get into AI without Python?

[LOCKSUIT]

They do go python, out of all my article etc readings over 3 full years of research i never seen C++ tutorials lol. Coursea and Hinton the AI pioneer all uses python.....but do look on coursea to see if there's c or c++ maybe there is...

But it's harder and more time consuming to code in C++....thats why they go python for Machine Learning.

BUT C++ will be efficient for speed/mmem size, so do it once u know....once u got it ready.

I pay coders for work, so, i went python, its like 25USD/per hour instead of 50USD/ph



As someone quite in against the ANN AI field way of how nets should work, and full 3 years i've been researching, i say for AGI you shouldn't have to worry at all and it scales, you can use a home pc. And for a adult mind still a home pc i say.

All others that do AI like are tought to, will tell you you may need a big google warehouse pc to make the full grand AGI.

There is Google Cloud, btw....however far that lets you idk totally.

[Art]

Welcome, Tempted! Nice to have you with us.

I'd think twice or thrice before storing ANY of my hard-earned programming, research, notes, experiments, etc. I simply have such a strong aversion toward and distrust of ANY CLOUD based location! Keep your research Locally, on your hard drives and storage devices. There can be no thievery there.

As I mentioned to LOCK, be sure to get a Non-Disclosure Agreement drawn up and signed Before entrusting anyone to collaborate, co-research, assist or program for you. If your project is of any importance to you then you should take proper precautions to safeguard it. Otherwise, why bother?

There are others here, far more learned than I, with regard to programming languages that would best suit your endeavor.

Best of luck!

- Art -

[8pla.net]

It costs beer money, close friends can chip in on.  I started making this for show at a party.

[tempted]

Thank you LOCKSUIT, Art and 8pla.net for your encouraging words.

[Korrelan]

It's not expensive these days to build a nice powerful mobile/ portable...



Note the wheels...

 

[8pla.net]

A big welcome to the author of this thread! Anyone interested in hearing about how to build an A.I. such as a chatbot, on a supercomputer?  Korrelan, first... Thanks for posting this. Please generally chat about that head node and the rest of the eight node cluster on wheels.  What's the fewest number of nodes needed for supercomputer building practice?

[WriterOfMinds]

Hi tempted, and welcome to the forum.

No one knows how to build a general AI yet, so nobody here can tell you what its requirements are.  If you hope to work on the problem, I think your best bet is to look at some surveys of possible approaches, decide which one(s) you are interested in trying, and come back with followup questions.

I run my own project (which I call a semantic net AI, though it's steadily becoming a kitchen sink of techniques) on my personal computer.  Chatbots like 8pla.net's should also be pretty undemanding.  If you want to run a power-hungry neural net, then you could be better off renting time on a cloud service.  (I feel Art's warning about this may be a little misplaced. If you actually invent something revolutionary, then it applies. But for purposes of getting into the field, the cloud is fine; nobody wants to steal the algorithm you wrote for a My First Neural Net tutorial.)

Almost all resources are suggesting Python, is there no way to get into AI without Python?

There is (you can write any program in any sufficiently complete language), but why would you want it? If you had not mentioned AI and had merely asked for recommendations as someone new to programming, I would have suggested Python anyway. I cut my teeth on Java and C, then switched to Python after deciding it was much less of a hassle to work with than either of those. More specialized languages may lend themselves to use on some AI research paths, but again, that depends on your desired approach.

Math: Try to develop a broad background in higher math -- enough so that you can recognize what branch of mathematics would be good for solving a given problem.  Unless you plan on writing a rule-driven AI that does not do any learning or data processing, I suggest a study of Probability and Statistics.  Further specifics would, yet again, depend on approach and task.

[LOCKSUIT]

Super important post:

The most important thing to know is that you can figure out how AGI works without doing any testing/coding. [First] figure it out, but make sure the [design] is efficient and not just Brute Force. However that will come on it own IOW just figure out AGI - recognition/generation depend on validation of claims i.e. hierarchical relations clustering - not brute force. Why do you think you can skip brute force? You can skip it tons (however not fully in some cases; you can't skip lock combinations (unless you have knowledge from the planet it's from)).

AGI=efficiency, there's little memory/processing concern.

Data in data out, you will discover new knowledge based on what you know and what you desire. AGI scales. Data is more data. It's just knowledge. Text takes up space accordingly. Vision though will take more memory/processing.