A Personal Blog
by Michael Froomkin
Laurie Silvers & Mitchell Rubenstein Distinguished Professor of Law
University of Miami School of Law
My Publications | e-mail
All opinions on this blog are those of the author(s) and not their employer(s) unelss otherwise specified.
Who Reads Discourse.net?
Readers describe themselves.
Please join in.
- Michael on Didn’t Waste Much Time
- Just me on Didn’t Waste Much Time
- Truth Pants on Didn’t Waste Much Time
- S on A 2024 Freedom Agenda (ver. 0.1)
- C.E. Petit on Does Tripit Think I’m on a Watch List or Financial Sanctions List?
Subscribe to Blog via EmailJoin 2,772 other subscribers
Peek Behind the Curtain
Today’s Deep Learning Is Like Magic – In All The Wrong Ways, is making the blog rounds and for a good reason — it’s a quick distillation of some essential truths about “AI” aka “Deep Learning” that the public needs to hear.This essay in Forbes,
This article is absolutely correct.
Unfortunately, even if it is believed by the non-computer-literate, it won’t likely be applied beyond the narrow scope of this article.
ALL computer code is simply a distillation of a human decision. All of it. Other than (arguably) a math problem (add X to Y), all of it is basically, “examine X, if it is equal to Y, then do this, if it is equal to Z, do this.” In every computer algorithm, there is the necessary human part of deciding the parameters and what to do about them. No computer decides that, because no computer CAN. Just as in that article.
We use waaaaaay too many “computer models” in every day life and I think, like the AI at the basis of this article, we elevate these models to a level far beyond their actual value.
The biggie here is “climate change.” The hard fact is we don’t have anywhere near the data that would be required to make a real scientific judgment of it. So the “scientific” community has manufactured data by using computer models, then using those results, fed into other computer models, to predict the future.
We forget that all any of this is is a very fast iteration of some individual (or group’s) reasonable belief on the matter. Because this belief seems reasonable, and because the result seems reasonable, it becomes fact, predicted by the models. But it’s not. It never was, because the data, also the product of models, never really existed. it comes down to an act of faith, disguised as science.
Now, I know, Michael, that you are a believer in all this, but I’d like you to take one step backwards from yourself and consider whether this article about AI, is perhaps (and maybe inadvertently) also an article about our modern world’s too much faith in something they don’t really understand?
So what evidence would you take as proof of climate change? CO2 levels? Temperature patterns? Glacier melt?
Honestly, I’m not sure we have enough evidence to garner together into proof. The vast majority of the evidence does not come from actual measurements, but from a computer program (codified supposition).
This should be obvious to anyone who thinks about it.
Ask yourself: How many years have we had weather thermometers that could measure to the accuracy necessary under the rigors of science? Remember, you need accuracy to a tenth, if not a hundredth, to be able to use those figures, you can’t claim accuracy that isn’t there in the measurement, as we all learned in high school science. And I’m talking weather station deployment, not existence in labs.
Now as yourself: what geographic area of the Earth has been covered by such accurate weather measurement during this period of availability? Do we have accurate measurements going back even 50 years for any significant percentage of the Earth? Do you think that’s even possible? (Remember, such measurements would have to have been considered important enough to make in the first place – why would anyone have cared about the temperature, to a tenth or hundredth, in some random wild corner of the planet in 1925?)
“Science” has made up for this obvious complete lack of measurement coverage by using computer models and “proxies,” (which are really the same thing). Do you really think there is any possible way to accurately know the wholistic climate of the Earth even 50 years ago, much less 500?
We simply cannot claim to have evidence that doesn’t also suffer from being massaged by human supposition. This is not science.
Does this mean climate change isn’t real, however you want to define it (man-made or natural)? No it does not, but we also can’t claim it as scientifically proven, because it is not. We cannot claim it to be happening, much less ascribe a cause.
And modern individual measurements of this or that, really mean nothing without both accurate direct past measurements AND an understanding of the entire system. If CO2 is higher now, how do we know it matters, other than supposition and computer models?
I’m really not trying to be contrarian. Your cited article laid it out. A computer simply cannot reach scientific conclusions, or any other conclusion, on its own, with the exception of math type problems that have objective single answers. You know this, and you must have seen this echoed in the article, or you wouldn’t have cited it as an “essential truth.”
I wonder, what’s your position on evolution?
How about your position on actuarial science?
Btw, I sort of agree with you. But, it seems to me that the scientific community is using their best data and their best methods to arrive at their most likely conclusions. That’s not to say the conclusions are hard fact (like the fact that I have a cup of coffee on my desk that I can touch and feel and know that it exists). But, those conclusions should be given some attention, particularly when there appears to be such wide spread consensus in the community actually dedicating itself to this research.
Put another way, what are the rough odds that the conclusions are right? What would it cost us if those conclusions end up being right? What would it cost us to act on the conclusions and to avoid them? Is the cost of prevention greater or less than cost (adjusted for likelihood) of the damage that we would suffer if the conclusions are correct? Using an actuarial model, it seems like climate change is “real enough.”
Even if the economic didn’t play out, to write off climate change altogether (and that may not be what you’re doing) seems to require a whole lot of faith in a conspiracy that involves thousands and thousands of people – and not a single defector, that I am aware of, has ever tried to shed light on that conspiracy.
We’ve gone a bit off topic. But I am curious as to your thoughts on this.
You need to really absorb what the article Michael posted is saying, and realize that it doesn’t just apply to things openly called AI. This is just how computers work. All the time. With every program. The are really just the automatic application of suppositions.
I don’t think these scientists are being dishonest in any conspiratorial way, but I think a lot of them haven’t absorbed this simple fact either. It’s not science in a traditional sense if you are simply asking someone or something else to confirm your suspicions for you.
We KNOW we don’t have the actual data needed. So it MUST be manufactured by some process. If Michaelks article is telling the truth, and it is, then why suspend that view just because lots of people are in agreement to something different?
The cost of “fixing” mad-made global warming are just too high to be wrong. How much harm to the poor will result in spending tens of trillions of dollars on something that neither matters nor can be fixed anyway? How many will starve needlessly so that people might not lose expensive real estate on Miami Beach? (Which I grew upon and which has been flooding like it sometimes does for my entire long life.)