🀷 Automated Death

Killer robots! | No more sun! | A new AI ethics institute!

🀷 Automated Death

Hello! Wanna hear something dumb? There is a cryptocurrency called Omicron and last week it was up 900% πŸ₯².

Anyway, despite that, this week was sort of okay I guess 🀷. Why? There's one good thing mixed in with some standard bad things. If you're smart, you'll know which is which:

  • Timnit Gebru has launched her own cool institute. Let's learn about what it will do!
  • The US government don't like giving kill orders, they want robots to do that instead...
  • Amazon runs a climate prediction model in their cloud and it spits out some harrowing results

πŸ’£CTRL-ALT-EXPLODE

Ah, war. The perfect mechanism for figuring out who in society is strong, and who is weak. Like with just about everything, the only thing that's stopping war from realising its full potential is the lack of automation.

The US government, lover of drone strikes, have decided that regulating the use of lethal autonomous weapons is... not a good idea. Instead they say that "the best way to make progress ... would be through the development of a non-binding code of conduct."

FYI: we are talking about the kinds of weapons that do not require a human to tell them when to do killing. I just think it's SO FUNNY that people like me sit at a desk on a daily basis, talking about the injustices of algorithmic bias, when there are pieces of tech out there that are literally designed to kill people, and even the UN are struggling to regulate this.

Personally, I think the 'best way to make progress' is to stop killing anyone (except Nazis, they can all choke on tar/drown in oil/be fed to our pets). Sure, I know basically nothing about regulating weapons, but what is a 'non-binding code of conduct' going to do? Does't that have about as much power as a 'don't do coke in the bathroom' sign?? Honestly...


πŸ™…β€β™‚οΈ Stop harming people with AI, I DAIR you...

Finally something good: Timnit Gebru (AI researcher fired from her position at Google a year ago for saying that large language models are bad) has now formed her own institute: The Distributed AI Research Institute (DAIR).

So, as you probably already guessed, DAIR will attempt to put an end to the harms of AI, by advocating for diverse perspectives inside of it's production. Taken from their website: "With DAIR, Gebru aims to create an environment that is independent from the structures and systems that incentivize profit over ethics and individual well-being." Broad stroke: Gebru wants to be able to do her valuable research in peace, without getting shot down by loud, profit-greedy idiots.

Okay great β€” but how? Everyone I write for, everyone who's work I edit, and everyone who I am friends with in the space all say that they want to do something like this. None of these ideas are that new; mostly I would say they stay as ideas and hardly ever turn into effective action.

I am naively hoping that DAIR will be different, probably because I just love setting myself up for disappointment. Key DAIR details that sound like they may lead to effective action:

  • DAIR has secured $3.7 million in funding: Gebru will in part use this funding to give her researchers space to work like humans, and not like academics. Academia is riddled with long hours and shitty deadlines β€” very glad to hear she's avoiding this model
  • She's not planning on living under an investor's foot: because investors seem to CARE what their money is doing (πŸ™„), and therefore expect you to take their opinions on board. So, Gebru wants to eventually find ways to sustain DAIR via consulting. Hope that works, and that she doesn't end up consulting for Clearview...
  • She's going to make the research accessible: I'll believe that when I see it, but apparently there's going to be effort put into ensuring the research with is shared and understood by affected communities. I think that's what the 'distributed' bit is referring to in DAIR.

This sounds promising, and I think Timnit Gebru produces interesting papers, but I'm wary of someone who is both as smart as this, and has assumed that they can change Google's policies from the inside. This quote is from this WaPo piece about DAIR:

Even as the high-profile co-lead of Google’s Ethical AI group, Gebru said she was more successful at changing Google’s policies by publishing papers that were embraced externally by academics, regulators and journalists, rather than raising her concerns internally about bias, fairness and responsibility.

To me, this fact is so obvious that it should go without saying by now β€” Google will never change their policies just because they pay you a huge salary and you have a cool job title.


πŸŒ‡ Stop having fun sun

Does anyone remember that Simpson's episode where Mr Burns tried to block out the sun? Good news! Maybe that will happen in real life!

There's a non-profit called Silver Lining who focus all their research on solar radiation management. Basically: how do we make the Earth less hot? Their answer: probably... dim that giant ball of fire in the sky?

🧊 Cooling facts:

  • Silver Lining have used Amazon's cloud computing platform to run their prediction model. This is apparently one of the first times a climate model has used cloud in this way β€” normally these models are run on a local supercomputer, so this a pretty big step in cloud computing capabilities.
  • Their model spat out 30 simulations of the Earth by the middle of this century. Some of them showed that humans went down a geo-engineering route, and kept the Earth cool by dimming the sun
  • We've already seen this happen in real life (if you assume that we AREN'T living in a simulation β€” be closed-minded if you like); volcanic ash clouds dim the sun for a long time, and therefore have actually cooled the Earth temporarily.
  • I didn't know this, but earlier this year, Bill Gates got on his high horse about geo-engineering, and is going to run tests where he blasts a chalky substance into the atmosphere or something.

But the coolest fact is: geo-engineering is scary, and like all technology, has unknown and therefore potentially unprecedented effects β€” but hey, that's never stopped anyone before!

Okay, that's enough for this week. Please, as is tradition now, have a Christmas party with wine and cheese and then deny it later β€” you deserve it.

Georgia


πŸ’Œ Seeing as you're at the end of this week's Horrific/Terrific, I guess you enjoyed it. Please express gratitude by donating money to me so I can keep doing this. You can share opinions/submit news stories any time on my Twitter or email.