2023-08-06

2023-08-06 Sunday - My criticisms of Yann LeCun's recent MIT presentation

Today, Dr. Yann LeCun posted on LinkedIn, about his recent talk at MIT ("Objective-Driven AI: towards AI systems that can learn, remember, plan, reason, have common sense, yet are steerable and safe", link to his slides)

My 1st comment:

"The configurator configures the agent for a deliberate ('conscious') tasks."

My suggestion:
A governance function (named in his honor, "The LeCun Inhibitor Function"?) that will prevent any tasks performed by an AI, from causing harm, or by omission, allowing harm to occur.

"Objective-Driven AI systems will be made subservient to humans"

How, without a governance function?

"Disinformation, propaganda, hate, spam,...: AI is the solution!"

That will only work, if you have a governance function.

"What if bad people get their [hands on] powerful AI? Their evil AI will be inferior to the Good Guys’ AI police." [see slide #68]

That right there illuminates the problem that I (and, I suspect, many of your critics) have with your position on the very real potential risks of AI causing harm. You, sir, are in denial.

Your view is grounded in some idealistic and ***very*** Pollyanna-type assumptions.

Police, in the real world, are not a preventative prophylactic - they are an after-the-fact attempt at error correction (and almost always insufficient to undo the harm committed), in almost all cases.

The implication of your statement is ALSO a totalitarian control of all AI activity.
Who will perform and regulate that function, for the entire world? Meta? To believe that, would be delusional.


My 2nd comment:
Based on LeCun's statements that I have cited from his presentation, my analogy:

He is driving a massively powerful (and very dangerous) race car - but he is ignoring the posted speed limits, caution signs, and driving very aggressively to reach his destination - no matter the risks. When told that he needs to slow down in the curves, he wants to accelerate. When told he needs better brakes, before increasing the horsepower, he effectively says, "we'll add them later".

The keys should be taken away from him, and his license to drive revoked. 🤣

No comments:

Copyright

© 2001-2021 International Technology Ventures, Inc., All Rights Reserved.