Scientists use AI to discover new antibiotic

[Leviathan-]

All I want to know is which one of us is going to make a client like FaH to discover new drugs. Strictly for scientific purposes.

[RejZoR]

God I hate this buzzword. Ai this, Ai that, Ai everything. And neither is actual Ai. Ai is when code makes sentient decisions and figures out problems on its own like we do. Telling it what it's suppose to do and then it basically brute force loops and trains itself within those parameters is not fucking Ai. Even all the "amazing Ai" presentations where computer figures out how to do certain task are bullshit. None of it ever just learned anything out of thin air, they had to tell "Ai" it's a humanoid shaped thing that's suppose to walk and jump over obstacles. No "Ai" just on its own figures out it's humanoid shaped object that can walk and jump the way humans do. I think Google had one such presentation. It's just ordinary algorithms we've been using for years and decades, they just store findings so they can use them in the future instead of doing everything all over again every time even though you found better output along the way without having to code the routine by hand to make it better.

 

None of this is "Ai". All of this are regular processing algorithms with store buffer that feeds the training database for future use. And that's it. God damn, I was far more forgiving when enemies in games were called "Ai" even though they had most basic routines to follow and attack the player and they learned nothing with time as you played against them. But the way how everyone is throwing "Ai" around these days is just cringeworthy to its most extreme possible extent.

[straight_stewie]

On 2/24/2020 at 7:56 PM, Shreyas1 said:

Seeing the high costs of medicine these days, hopefully this will bring costs down for many as developing these drugs will hopefully  now be far cheaper and easier. 

Something tells me that the cost of developing or manufacturing a drug has little to do with the consumer cost of that drug in most cases.

[straight_stewie]

29 minutes ago, RejZoR said:

It's just ordinary algorithms we've been using for years and decades, they just store findings so they can use them in the future instead of doing everything all over again every time even though you found better output along the way without having to code the routine by hand to make it better.

 

29 minutes ago, RejZoR said:

None of this is "Ai". All of this are regular processing algorithms with store buffer that feeds the training database for future use.

What?

That's not even close to how a neural network works. Like, at all.

Maybe you could make the argument that a Genetic Algorithm just produces "ordinary algorithms we've been using for decades". Maybe. But other than that one, small, very narrowly scoped caveat, nothing about what you just said is based in fact.

[RejZoR]

Do you seriously believe I'm gonna go in deep details for a forum post? Where's logic in doing that?

 

If it is called machine learning, or ML for short I wouldn't even have problem with it because it's exactly that. Machine learning. It's no Ai and it won't be for decades if not longer.

[Engineer.AI]

5 hours ago, RejZoR said:

Do you seriously believe I'm gonna go in deep details for a forum post? Where's logic in doing that?

 

If it is called machine learning, or ML for short I wouldn't even have problem with it because it's exactly that. Machine learning. It's no Ai and it won't be for decades if not longer.

Agreed. Although it's called AI, there is little that bears semblance to what, for example, Asimov wrote about in his works. People should be educated on the difference between AI, ML, and DL - only if for the sake of literacy. I also loved your earlier post, where you call artificial intelligence a 'buzzword'; it really puts the entire stream into perspective.
~Engineer.AI

[straight_stewie]

14 hours ago, RejZoR said:

It's no Ai and it won't be for decades if not longer.

AI is classified into three categories.
 

• Narrow
• General
• Super

Turns out that Machine Learning, as we know it today, is part of the definition of the Narrow category...

Unfortunately not every word or phrase gets defined as the most obvious thing. It would make English quite a bite easier if it did, but what can you do? You can make all the arguments about it not being a General or Super AI that you want, but at the end of the day the commonly understood definition of AI can be distilled to "a program which can learn from new data and which can solve problems that it was not directly programmed to solve".

Given that common definition, the antibiotic discovery application we've seen here falls into the Narrow category of Artificial Intelligence.

But I totally agree with your larger point: Too many people have too great a misunderstanding of what AI is or what it's capabilities are. I would also agree with your point that changing the commonly understood definition of "Artificial Intelligence" would help curtail these misunderstandings, but with the stipulation that it is a losing game to try to change the definition of words that the scientific community has reached consensus on.

[Syaoran]

We don't need new antibiotics, we need bacteriophages!