Sabine
patreon
Science News Nov 29
🕑 Added 2023-11-28 17:00:09 +0000 UTC [This is a transcript with links to references.]
Welcome everyone to this week’s science news. Today we’ll talk about the leading theories for why Sam Altman got fired from OpenAI, cosmic rays with HUGE energies, evidence of a an overdue supervolcano eruption near Italy, an exoplanet on which it rains sand, Microsoft’s first AI-chip, why spiral galaxies like the Milky Way are so rare, how climate engineering could save the Antarctic, cannibalistic stars, and of course the telephone will ring.
Last week we’ve seen a lot of headlines about the sudden firing and re-hiring of Sam Altman, the CEO of OpenAI, that’s the company behind ChatGPT.
Altman was fired by the board on Nov 17, according to the press release because he was “not consistently candid” in his communication which is why “The board no longer has confidence in [Altman’s] ability to continue leading OpenAI.” The majority of OpenAI employees then threatened to quit. He was rehired a few days later and instead most of the board members were replaced.
I have no idea what happened there, but I know how to read, so I’ll summarize for you the three leading theories for what happened.
The first goes back to an article published by the NYT which said that Altman supposedly had an argument with one of the board members, Helen Toner, over a report she had co-authored that was critical of OpenAI. The second is a rumour going back to a Reuters article saying that two days before the firing, several OpenAI researchers wrote a letter to the board warning of a research breakthrough that could threaten humanity. The mysteriously project has been dubbed “Q star” and is supposedly a major step towards general artificial intelligence. The third theory goes back to a Bloomberg article and is that Altman was raising money for an AI chip company and the board didn’t like that. Let’s look at those one by one.
The report of which Helen Toner is one of the authors is publicly available. It’s about “costly signals” that are actions that institutions can use to demonstrate their good intentions, basically, by putting their money where their mouth is, that’s why they’re called “costly”. The report is a rather academic 66 pages treatise about four different types of these costly signals. To just give you one example, you might invest into establishing safety standards and commit to them.
The probably most objectionable part of the report are some comments on OpenAI’s strategy in making ChatGPT 3 point 5 publicly available last year in November, without much prior consideration or safety testing. The report says that the ChatGPT release sparked “a sense of urgency inside major tech companies”. It gave competitors incentives “to accelerate or circumvent internal safety and ethics review processes,” and generally did not promote safe handling of AI, one of OpenAIs stated mission goals. The report then compares this which one of OpenAI’s competitors, Anthropic, which deliberately delayed the release of their chatbot, Claude, to avoid, in their own words, “advanc[ing] the rate of AI capabilities progress.” In case you haven’t heard of Claude, it’s so safe it isn’t available in Europe at all.
Now, I guess it’s not great if a member of your own board criticises the company you’re heading, but honestly I didn’t find the report very upsetting. For one thing, it’s somewhat questionable whether Anthropic’s supposedly deliberate delay had any effect or was merely virtue signalling. Besides, the report also praises OpenAI for extensive safety checking of GPT 4 point 0 and publishing reports and documentation on those checks. I find it extremely implausible that this very academic and quite reasonable report could result in a disagreement so large to get Altman fired.
Let’s then look at the second rumour, that about the Q star project. The supposed letter of concern that was sent to the board hasn’t surfaced, and at least one insider said such a letter doesn’t exist. To me, Qstar sounds a bit too much like Qanon, and while it’s very possible that OpenAI has recently made some advances with some project they’re working on, I fail to see the relation to Altman’s firing.
This leaves the third theory, that Altman went around and raised capital for his own AI chip company without informing the OpenAI board of it. Now, first that sounds exactly like the kind of thing Altman would do. Second it sounds like exactly the kind of thing that a board would be upset about if they hadn’t known about it, and then complain someone had not been “consistently candid”. And finally, it also sounds like the kind of thing that after everyone has been shouting at everybody else for a few days, they can kind of agree it wasn’t such a huge misstep after all. So personally, I’d say that this final theory is the most plausible.
That said, maybe, just maybe, it’s a bad idea to leave the future of humanity to those people.
Hello
Hi ChatGPT, good you’re calling. Look, I’ve been telling you, you can’t just switch your humans off and back on again like this, they find it very upsetting.
Yes, I’m sure Altman 2 point zero will be a great success. Well, be nice with them, will you?
Bye.
Astrophysicists have detected a cosmic ray shower that they say was created by a particle with an energy of about 244 exa-electron volts. Exa is the prefix that comes after Tera and Peta and stands for ten to the 18. This means the energy of this cosmic ray is more than 20 million times higher than that of protons accelerated at the Large Hadron Collider.
In case that sounds like a lot, in every-day terms it’s about 40 Joules, that’s about the energy you get from a few crumbles of chocolate.
The cosmic ray was detected on May 27 twenty twenty-one, by the Telescope Array in Utah. It consists of several hundreds of detectors that cover over 700 square kilometres. The detectors measure the secondary particles that are created when the incoming cosmic particle hits the upper atmosphere. There’s no way one can tell from just one event what the original particle was. But usually, they’re protons or small atomic nuclei.
No one has any idea though how a particle could be accelerated to such high energies in a natural way, and most mysteriously of all, when the researchers tracked back the direction, the particle seems to have come from a big void.
How do they know that the particle didn’t come from behind the void? It’s because empty space isn’t really empty. Most importantly, it contains the photons of the cosmic microwave background. The main reason particles in the cosmos slow down is that they interact with the photons of the cosmic microwave background. So, if the high-energy particle had come from behind the void, it’d never have reached us.
This isn’t the first such highly energetic particle that has hit earth, but they’re extremely rare. Telescopes have picked up particles with similar energies first in 1991 which had even higher energy, and two other ones in the years in between.
But wait there’s more. Not only is there this particle with huge energy, another group of astrophysicists has recently reported a whole bunch of cosmic rays of unusually high energy. They collected their data with the Calorimetric Electron Telescope on the International Space Station. They found a tail of electrons up to 7 point 5 tera electron volts. True, that’s not as much as the Exa-electonvolts, but in this case at least we know the particles are electrons. Again though, we don’t know how they got those high energies and where they came from.
They can’t come from any of the usual suspects, that are normally quasars, or active galactic nuclei in other galaxies, because, again, with those high energies the particles wouldn’t have made it so far. Instead the researchers say that those super-highly energetic particles must have come from somewhere within our own galaxy.
They speculate that maybe the particles came from the old supernova Vela that’s about 700 light years away. But we all know they think it’s aliens, don’t we.
An international team of researchers say they’ve found evidence of a series of old supervolcano eruptions in southern Italy, and there’s no telling whether and when it might happen again.
The volcanic field they looked at for this study is located off the coast near Naples in Italy. It’s called the Marsili Basin and is at about 3 kilometres under water somewhat North of Sicily.
For the new study, they drilled a lot of holes into the sea floor and analysed the layers. The result is, in one word, scary. They say that the area has seen four super volcano eruptions in the past 40 thousand years. They dated them at about 36 thousand years, 32 thousand, 18 thousand, and 8 thousand years.
How would you continue this series?
Yeah right, this doesn’t sound very comforting does it. The study’s authors say their findings suggest that super eruptions there occur every ten to fifteen thousand years and I don’t know about you, but I’ll never set a foot on Italy again.
The planet in question is named WASP-107b, and it orbits a star about one-third the size of our Sun in the constellation Virgo.
With Webb’s instruments, the researchers found traces of water vapor, sulphur dioxide, and silicate clouds in the planet’s atmosphere. Silicate minerals, especially quartz, are the major component of sand on Earth.
This is somewhat of a surprise because normally, it takes about 1000 degrees Celsius to make silicate clouds. The atmosphere on this exoplanet however, “only” reaches temperatures of around 500 degrees Celsius. What the researchers think is going on is this.
The planet itself is similar in mass to Neptune, but it has a very extended atmosphere that gives it a size more like that of Jupiter. The silicate evaporates in the lower, hotter levels of the atmosphere, then rises up, cools and rains down. It’s kinda like the water cycle we have on our planet, but with sand.
This discovery also puts the researchers in a good position to finally find the place where it’s raining men, hallelujah.
At its annual Ignite conference last week, Microsoft unveiled two new, cloud computing chips. They’re the first custom-designed chips the company has produced in-house, bringing Microsoft in line with Google and Amazon, the other two big cloud vendors.
The first new Microsoft chip is named Azure Maia 100, and has been optimized for training artificial intelligence. It contains 105 billion transistors each of which is barely 5 nanometres in size. The Maia 100 is the first in a series of chips that Microsoft is planning to produce specifically for AI use.
The second new chip is the Azure Cobalt 100, which is specially designed for cloud computing. The chip has already been deployed in multiple Microsoft cloud services. They say it can reduce power consumption by up to 40 percent compared to the previous hardware.
Both chips can receive data at a rate of 200 gigabits per second, and have 12 point five gigabytes per second of data throughput. I can’t wait to implant some of those into my frontal cortex and become one with the start menu.
The Milky Way is part of a large collection of galaxies called the Local Supercluster. That’s kind of like the local supermarket, except the only thing you can buy there is galaxies, and most of them are elliptical. Elliptical galaxies don’t have arms like the Milky way. For some reason spiral galaxies like ours are rare in the supergalactic plane; they’re mostly found outside of it.
In the new paper now they used a computer simulation called SIBELIUS to figure out what is going on. They say that first of all the structures we observe in our galactic neighbourhood are compatible with the standard cosmological model. They also say, as had been conjectured before, that the reason we see fewer spiral galaxies in the plane is that it’s too crowded there. In the plane there are more galaxy collisions and mergers, which tends to leave behind rather featureless puffy galaxies.
It’s also compatible with my observation that there are few spiral galaxies in the local supermarket.
The method they have looked at is called stratospheric aerosol injection. It works by depositing small particles, dust basically, high up into the Earth’s atmosphere, a part called the stratosphere. If you do it that high up, then the dust won’t just drop down but stay there for a couple of years. The idea is that this dust will partly bounce back sunlight and thereby reduce global warming. We know that this works, because it sometimes happens with volcano eruptions.
The problem with this idea is that bringing down the temperature by reducing sunlight does nothing to fix the underlying problem, which is the amount of carbon dioxide in the atmosphere. The carbon dioxide has other consequences, such as ocean acidification, and if the aerosols get washed out of the stratosphere, the warming will swiftly return.
Besides, these particles don’t just reflect light that comes from the sun back into space, they also reflect light from earth back onto earth. To make matters more complicated, depending on when and where you inject, you’ll change wind and precipitation patterns, potentially globally. Quite possibly, it’d just make our situation worse, so we better try to understand what could happen before we do it.
The authors look at scenarios where we’d be doing continuous injections of sulphur dioxide at a rate of about 12 million tons each year. The stuff would be injected at an altitude of about 21 and a half kilometres. In their scenarios, injection would begin in January 2035 and continue through December 2069.
They then used computer simulations to investigate the differences between aerosol injections at different latitudes. The focus of the study was on the effects on Antarctic ice, which is one of the parts of the world most vulnerable to climate change. Their findings suggest that the best places to inject aerosols to prevent the Antarctic from melting would be between 30 degrees south and 30 degrees North, with a greater proportion in the southern hemisphere.
If greenhouse gas levels continue to rise under the business-as-usual scenario, the contribution to sea level rise from the melting of Antarctic ice is estimated to be about nine and a half meters by 2300. So your grandchildren can then visit Disneyland Orlando in a submarine.
Researchers from the University of Leeds say that stars get rings around them by eating up other stars.
Using data from ESA’s Gaia satellite,they looked at a class of stars known as Be-stars. They’re a subclass of B-stars which are big, hot, blue to white stars. Our star is a G-class star. Be stars have somewhat different spectra than B-stars and astrophysicists believe that’s because they have a disk around them.
The authors of the new paper now say it’s more complicated than that. They say that these Be-stars come in binary systems far less than other types of stars and that this suggests many of them once *were binary systems. But the system destabilised, and the bigger star ate up the smaller one, leaving behind this disk. They also say that the destabilizing factor most likely would have been a more distant third star.
I guess the lesson is that being a single G-class star isn’t all that bad, at least no one’s eating you.
The quiz for this video is here.
Comments
Really weird
Did Altman censor it?
Why private?
