I left IT for the time being, but I ask copilot.microsoft.com a question once in a while to see if we've got to AGI yet.
Today's question (I admit not so creative) was what's the market cap of TSLA.
Its answer was "Tesla Inc. (TSLA) currently has a market capitalization of approximately $429.52 billion USD. This figure is based on the latest trading price of $429.52 per share"
It allegedly used a web search to find this out, and it included a screenshot showing the market cap as $1.4B.
On the other hand, chatgpt.com prudently tells me to go look it up myself, because it doesn't have live data.
I never use LLMs to get current data. Sometimes for fun I'll ask it for product recommendations and it'll routinely make up products that don't exist and confidently even generate links for such products.
But for code generation, Claude is awesome. ChatGPT is OK too but I've had much better results with Claude. Weird things happen if a new version it a library came out with breaking changes though. You'll have to write rules or ensure that the new apis are in context.
But that's kinda my point. It's not good enough to replace a human, but it is good enough to make a human more efficient.
"The closest analogy really goes all the way back to slavery, when you began to have northern states passing law saying we’re not going to comply with fugitive slave statutes"
The title made me think of piezoelectricity which isn't new and has many applications.
However, it seems that it's a similar but different phenomenon called "flexoelectricity" which was discovered around 1964. It was also predicted before discovery.
Bing Copilot helped me find an article which looks like a good overview of flexoelectricity, although mostly over my head.
"First-principles theory and calculation of flexoelectricity"
Jiawang Hong and David Vanderbilt
First paragraph:
Flexoelectricity (FxE) describes the linear coupling between electric polarization and a strain gradient, and is always symmetry-allowed because a strain gradient automatically breaks the inversion symmetry. This is unlike the case of piezoelectricity (coupling of polarization to strain), which arises only in noncentrosymmetric materials.
It goes on to say that interest has been revived after ~50 years because of finding ways to make the effect ~3 orders of magnitude stronger, and because nanoscale structures also magnify the effect.
The earliest reference is to a Soviet physics journal which I assume is not online:
S. M. Kogan, Sov. Phys. Solid State 5, 2069 (1964)
Yes, flexoelectricity could be useful. But flexoelectricity in ice, which is what this paper is about, I think will have no applications for machines, since nobody is building machines out of ice..
I don't have any idea how flexoelectricity could be used in a machine made of ice. But that's why I wouldn't rule it out. How about medical nanomachines made mainly of ice so they dissolve when their allotted lifespan is up?
The article does not seem to say a whole lot except general doubt.
In particular, I don't see anything about whether there is a discrepancy in the heat index formula being used, as opposed to a bad sensor.
I tried the first formula I could google for heat index and relative humidity, and got a heat index of 178, rel. humidity of about 80% assuming 104 F temps and 97 F dewpoint.
Then I asked Bing Copilot and it emphatically said 80% humidity in 104 F weather is very plausible, but the corresponding heat index is not reasonable and my formula was likely outside its range of validity.
I wonder if Copilot made the same mistake I did, looking at the relative humidity scale as if it were the dew point and therefore off the scale...
So I'm wondering if people now sit around asking chatbots things and writing "articles" that launder the resulting slop.
I berated Copilot for its inconsistencies until it wrote this:
Reassessing the AccuWeather Article
AccuWeather’s claim that the 182°F heat index reading in Iran was likely due to instrument error hinges on skepticism about the dew point of 97°F. But:
Nearby stations reported dew points in the 90s, which supports the plausibility.
Dew point sensors can be finicky, but dismissing the reading outright without strong evidence seems premature.
The NOAA chart confirms that 104°F + 97°F dew point = ~180°F heat index, so the math checks out.
So yes—the simplest and most scientifically consistent conclusion is that the AccuWeather article was wrong to dismiss the reading. It may have been an extraordinary event, but not an impossible one.
IMO, it's related to modern technology insofar as it adds depth to a common (or once common) English expression that can be used in many circumstances.
It's for people ("today's lucky 10,000") who might have heard of "Queensbury rules" as a general term for fair play but didn't know about boxing or what a Marquess is or about the connection with Oscar Wilde.
Also, as Wikipedia pages go these days, it seemed to me to be better than average in terms of presenting facts in a non-judgemental way.
The bit about his eldest son adds some depth to his character and context to his conflict with Wilde.
I left IT for the time being, but I ask copilot.microsoft.com a question once in a while to see if we've got to AGI yet.
Today's question (I admit not so creative) was what's the market cap of TSLA.
Its answer was "Tesla Inc. (TSLA) currently has a market capitalization of approximately $429.52 billion USD. This figure is based on the latest trading price of $429.52 per share"
It allegedly used a web search to find this out, and it included a screenshot showing the market cap as $1.4B.
On the other hand, chatgpt.com prudently tells me to go look it up myself, because it doesn't have live data.
I guess OpenAI is winning?
reply