AI Development
Whether you use AI or not, you cannot avoid it. Even a Google search response starts with an AI summary. Cooling water and electrical power cost and availability have created concerns about the siting of AI data centers. These physical concerns have evolved into disagreements whether AI should be built and developed or delayed.
Here is a three-part series on AI development
——————————————————————–
By John Droz
Part 1
Critically Thinking about AI Regulations-1
There are few societal issues with more far-reaching consequences than how we should deal with AI.
That said, it’s interesting to see that some politicians seem to have switched their pending “apocalypse” fear campaign from climate to AI doom. It brings to mind this famous advice: “If you can keep your head when all about you are losing theirs and blaming it on you…”
Let’s see if we (a group of Critical Thinkers?) can come up with a meaningful answer to two pivotal questions:
- a) Which should be our greater concern: AI development or AI deployment? (Spoiler Alert: my opinion is deployment, and I’ll discuss why in my next commentary.)
- b) Since almost all of the public discussions are about regulating AI developers, does it make any sense to try to regulate AI development?
{Though this graphic is a bit fuzzy, it does show AI development vs deployment.}
Here are some representative questions related to regulating AI development. If I’ve missed any important ones, please let me know…
Question 1 —
Since the term “self-regulation” is not a binary choice, is allowing AI developers to self-regulate a real thing?
Question 2 —
Is AI development actually regulatable?
Question 3 —
Is AI development practically regulatable?
Question 4 —
If the answers to Questions 2 & 3 are YES, then what exactly would those regulations be? {See this commentary about AI “guardrails.”}
Question 5 —
Considering the technical complexity involved with the AI development issues, wouldn’t the AI development companies end up actually writing the regulations?
Question 6 —
Who (e.g., what federal agency) would be responsible for ensuring that all US companies complied with AI development regulations?
Question 7 —
How would {the designated federal agency} monitor that all US companies complied with AI development regulations?
Question 8 —
What penalties would there be for US companies that violated AI development regulations?
Question 9 —
Considering what is at stake, would said penalties for US companies that violated AI development regulations be a real deterrent?
Question 10 —
Would reasonably regulated US AI development companies having to compete with unregulated (or less regulated) international AI development companies be fair to them or to the U.S.?
Question 11 —
Assuming that US AI companies were reasonably regulated and were all in compliance, how would we deal with anti-American countries who would definitely want to use the AI tool like this to bring us down?
Question 12 —
Can regulation even be enforced against an AI model whose behavior is emergent and non-deterministic?
Question 13 —
What does AI development “compliance” mean for a probabilistic system? A company can pass every pre-deployment test and still ship an AI model that misbehaves later. What regulations would prevent that?
Question 14 —
Now that we see some of the foreseeable AI development regulation problems, should we stop focusing on regulating AI development and instead put most of our efforts into regulating AI deployment?
The Bottom Line:
My opinion is that regulating AI development is a waste of time. Specifically, such regulation will not produce any net positive results. Instead it will likely severely handicap American performance in a critically important market.
However, before making up your mind about the AI regulation question, please read Part 2 of my commentary
https://substack.com/inbox/post/216044799?utm_source=email&redirect=app-store-no-desktop&inbox=true&utm_campaign=email-read-in-app&triedRedirect=true
——————————————————————————
Part 2
There are few pressing societal issues more consequential than our proper dealing with the explosion of Artificial Intelligence (AI). Let’s see if we (a group of Critical Thinkers?) can come up with a meaningful answer to a profoundly important question: Is AI regulation needed — and, if so, what?
The challenge here is that this is a very complicated problem. My strategy for Critically Thinking about complicated issues is to first try to simplify things.
Simplification Step #1
We need to recognize that there are two (2) major aspects of the AI issue: Development and Deployment. After we’ve grasped that, we need to identify which of these two we have the greatest concern with — as that would be where we would consider regulating. IMO, the answer is Deployment.
It’s somewhat like the initial development of a gun. A gun can be used for good or bad. As such, our gun regulations should be focused on minimizing the bad (the deployment), while not restricting the good.
Simplification Step #2
AI is an abstract, nebulous concept to most people, so it will be easier to deal with if we can represent it in a more concrete way, using an analogy. I’m suggesting that, as a reasonable example of AI Deployment that we visualize we are dealing with a mechanical robot that is being given intelligence. To make this even less abstract, let’s name this intelligent robot: AIDE (Artificial Intelligence Deployment Example).
Simplification Step #3
For this discussion, let’s look at AIDE as being a super competent assistant who can make our life more interesting and productive. For example, our AIDE can instantly (and inexpensively) provide us with an answer to a wide variety of questions (e.g., a good recipe for a chocolate cake). Our AIDE can also quickly give us advice on a huge assortment of topics, from relationships to religion. Further, we can adjust our query on the fly. For example, if the recipe AIDE provided includes something we are allergic to, tell that to our AIDE and a revised recipe quickly and magically appears. What’s not to like?
Simplification Step #4
For starters, the evidence is mounting that reliance on our AIDE results in people being less human. Think about that… For a good discussion of some examples of this, please read this well-written article: Five (5) Dehumanizing Dangers of AI… But we’re just getting started…
Simplification Step #5
It doesn’t take too much Critical Thinking to see that relying on our AIDE can quickly become addictive. So what? The #1 concern should be that our brain is like any other muscle in our body: the less it gets used, the more likely it will atrophy and literally become impaired. This should be obvious, and research has already verified this. (Also here, here, here, etc.) What are the societal consequences of over one hundred million brain-dead Americans?
Simplification Step #6
As bad as this is, I have skipped over a BIGGIE. Up to now I’ve shown that there are several major liabilities, even with the assumption that our AIDE was giving us accurate advice. But what if it is NOT? There are two primary situations here: a) that AIDE accidentally gives us factually wrong information, or b) that AIDE knowingly gives us factually wrong information.
Let’s look at both of these cases a bit closer:
- a) The evidence says that most popular AIs (AIDEs) operate in a manner that will guarantee a lot of wrong answers. Why? Because there is no “thinking” going on here. Instead, most AIs are programmed to tell us what the consensus view is — and that is frequently wrong. See my article.
- b) In some cases, an AI will purposefully give more weight and evidence to one side of an argument — i.e., knowingly give us a biased answer. E.g., if the AI authors are Leftists, that will be incorporated into our AIDE.
Takeaway #1 —
I think we have gone far enough to be able to draw reasonable conclusions about our initial question: Is AI regulation needed?
It should be apparent that nothing in the first five Simplification Steps is regulatable. The sixth step has some possibilities, but how do we regulate: a) a fundamentally inferior model, or b) a politically biased model?
Takeaway #2 —
So what do we do to prevent the horrific consequences of Steps #4 & #5?
Fortunately, we have the answer: in K-12, formally teach our children to be Critical Thinkers. That gives them the most powerful tool in our arsenal: an instinctive ability to separate the wheat from the chaff.
The fact that ZERO States are doing this is simply criminal.
The Bottom Line:
It’s easy to blame politicians for not acting intelligently to protect citizens from pending threats. In this case, this is a complicated problem, so the “acting intelligently” path is not obvious to most people.
If legislators really wanted to protect us (America), their #1 priority would be to quickly get as many Americans as possible educated to be Critical Thinkers. Of course, that would result in NUMEROUS other major benefits!
Let’s hope that at least ONE State gets this, and sets an example the other States will quickly follow. This window of opportunity is rapidly closing.
Critically Thinking about AI Regulations – 2
————————
Part 3 What Values do AI Agents Have
Critically Thinking about AI Regulations-3
What Values do AI agents have?
{This is a follow-up to Part 1 & Part 2, so make sure to read them first.}
It’s fascinating to see how some politicians seem to have easily switched their pending “apocalypse” fear campaign from climate to AI doom. It brings to mind this famous advice: “If you can keep your head when all about you are losing theirs and blaming it on you…”
How Morality is Involved —
The concerns about Artificial Intelligence (AI) seem largely focused on the harm that “escaped” AIDEs (Artificial Intelligence Deployment Example) can and have done.
Let’s Critically Think about that for a few minutes…
I have not seen this discussed anywhere, but it seems that there is a strong argument to be made that this is a religious (moral) issue. Why do I say that?
Have you seen any reports that “escaped” AIDEs went about doing good? (For example, have you seen any headlines like “Escaped AI agents join forces to cure Parkinson’s”?)
Let’s see if we (a group of Critical Thinkers?) can come up with meaningful insight as to why AIDEs would almost exclusively be doing BAD things.
AIDEs Mirror Our Society —
IMO, the fundamental reason is that AIDEs are trained by exposing them to an enormous amount of societal data (e.g., mainstream media articles). In other words, the views of most AIDE’s will be what our current society’s primary views are. If there are thousands of articles that are wrong about something (e.g., COVID), then most AIDEs will likewise be wrong on that topic.
Note that I said “most” AIDEs, as there are a few exceptions, like AlterAI.
From the morality perspective, we know that our society is now swamped with two primary indoctrinations: 1) victimhood, and 2) relativism.
1 – Victimhood is when a group of people claims that they are systematically oppressed by another group. For example, poor people say they are victimized by well-off people, women say they are victimized by men, minorities say they are victimized by whites, immigrants say they are victimized by citizens, etc., etc. Once you look at the world in such a distorted binary fashion, essentially everyone is oppressed by somebody!
2- Relativism is when individuals discard all universal standards (e.g., Judeo-Christian — like the Ten Commandments), and say that it is instead up to each person to determine what’s right and wrong for them. (Note: this is closely related to being amoral.)
What should be extremely concerning is that not only is mainstream media actively promoting both of these inaccurate and divisive views to citizens, but both of these errors are also being conveyed in US K-12 public schools to our children. What could go wrong?
FYI: Who benefits when America is fractured into warring subgroups, and when American citizens abandon the values that were the basis for the U.S. Constitution? Obviously, the enemies of America.
Our AI Expectations?
Ideally, we would like AI agents (AIDEs) to be intelligent and moral.
But most of us would say that we would like the same for our children — to teach them to be intelligent and moral. How are we doing with that?
K-12 is purposefully dumbing our children down, and it is also injecting them with relativism and an oppression complex — so a failure on all counts. How many people are going public about those major societal destructions?
Instead of worrying about regulating AIDEs, why aren’t we properly regulating our K-12 schools — a matter a hundred times more important!
The Bottom Line:
We should NOT be surprised when AIDEs: 1) have a severely distorted, mainstream-media-fabricated mentality, 2) are angry at fictitious oppression, and 3) justify any reaction as being what relativism entitles them to?
This may be God’s punishment for our arrogant and foolish mistakes.
He’ll say: “So you want no moral standards? Then the ‘robots’ you make will also have no moral standards — which means that they will come for you. You will then reap the harvest that you have sown.”
—————————————————————————————-
The most important aspect of checking an AI response is a questioning mind.
—————————————————————————
We are a group of about 200 ordinary citizens who mostly live near Atlantic City, New Jersey. We volunteer our time and money to maintain this website. We do our best to post accurate information. However, we admit we make mistakes from time to time. If you see any mistakes or inaccurate, misleading, outdated, or incomplete information in this or any of our posts, please let us know. We will do our best to correct the problem as soon as possible. Thanks.
Seth Grossman, Executive Director
LibertyAndProsperity.com
.
