Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
How many quarters after Q1 2016 till Marissa Mayer is unemployed?
1 or fewer
2
4
Her job is guaranteed; what are you even talking about?
View Results
 
  • Post
  • Reply
Morrow
Oct 31, 2010
The issue is training an AI to follow make and follow a strategy i.e. "I'm going to push for cultural victory" and then convincingly follow that strategy but also be prepared to abandon it if circumstances change. A big issue with chess bots, f.e., is they don't play human: they may have some scripted gambits they like, but fundamentally it's "optimal play but blunder XX% of the time". Part of how people can tell when someone is cheating in chess by relying on an engine is they don't seem to follow a cohesive plan, because a human will hold a strategy in their head that they work to advance while an engine will just take the most optimal move after running thousands of simulations.

Adbot
ADBOT LOVES YOU

Mega Comrade
Apr 22, 2004

Listen buddy, we all got problems!
I wrote a chess engine for my dissertation. It was good enough to repeatedly beat my professor who was an avid chess player. However my friend who could barely play chess sat down and won 2-1.

Eventually I figured out I had weighted the engine too heavily to assume the human player would play optimal (or close to) moves, so it struggled to predict my friend who was playing mostly at random. It was something I hadn't even considered when starting out.

Family Values
Jun 26, 2007


evilweasel posted:

the issue here is that a computer game AI isn't designed to play the game, it's designed to convincingly lose in a pleasing way that makes you feel very accomplished and smart. if the computer beats the pants off you, you get pissed off and stop playing.

so the trick isn't getting the computer to play it very well, it's getting it to play in a way that is perceived as well, but lets the player win. but it's very hard to tune that carefully for the long term so the computer keeps presenting a convincingly pleasing, but beatable, challenge as you scale in skill. especially since the developers may not have a firm idea of how high the skill levels can get.

Yes... but no. People play against other people in strategy games (thinking Paradox-style multiplayer grand strategy games) if they want a real challenge. The problem is that competent competitive play exposes how unbalanced those games usually are and that there's an optimal, not very fun or 'flavorful' way to win and anything else is a loser's strategy (like EU4 multiplayer you either go quantity or go home).

An actually challenging AI would require similar unfun play in single player games.

A way around it might be to have multiple AI personalities. Some that were trained for varying amounts of time, some that were trained in restrictive environments (only allowed to take certain choices and never others), and a few that are absolute beasts. Then which AI you get will determine your style of play.

SaTaMaS
Apr 18, 2003

Nothingtoseehere posted:

Only with full visibility of the map, iirc. Getting current AI models to work on limited information games has been very tricky so far iirc. There was one for Poker recently, but iirc it brute-forced all possible states and just played the odds better than any human can.

It had full visibility of the map in that it didn't need to scroll around, but it still had fog of war.

Elias_Maluco
Aug 23, 2007
I need to sleep

Clarste posted:

Didn't we literally have a computer teach itself to play Starcraft at some point? The tech is there, it works, it just takes investment in letting a computer play against itself for a million billion hours while it figures out all the rules. Scaling geometrically with complexity, which I assume means it would take forever to learn Civilization.

Edit: Trying to use any kind of learning model analogous to ChatGPT where it just studies human games and tries to copy them seems like a good way to make a very dumb AI that does the strategic equivalent of drawing too many fingers on every hand. Like, it'd set up a perfectly timed attack on a place with no city because it doesn't know what a city is, it just knows that people usually attack places like this.
I bet that dumb AI would still be s lot more effective than Civ 6 AI

edit: I mean, even if an AI was able to play remotely like a bad player with cognitive problems, that would be revolutionary today. Strategy games AI (até least for games like civ) play like a blind man throwing darts at random, and the only way for them to pose any challenge is by having like a thousand more darts than the player

Elias_Maluco fucked around with this message at 17:25 on Mar 10, 2023

Fate Accomplice
Nov 30, 2006




Today's tech nightmare:

https://twitter.com/jamiequint/status/1633956163565002752

PhazonLink
Jul 17, 2010
e for reference post

Mega Comrade posted:

I wrote a chess engine for my dissertation. It was good enough to repeatedly beat my professor who was an avid chess player. However my friend who could barely play chess sat down and won 2-1.

Eventually I figured out I had weighted the engine too heavily to assume the human player would play optimal (or close to) moves, so it struggled to predict my friend who was playing mostly at random. It was something I hadn't even considered when starting out.



that reminds me of a the episode plot from hiraku no go. the antagonist/ rival had to play a game behind his back and struggled with remembering what moves the youngest / weakest player did.

wood tier so dense it hits like a neutron star matter.

PhazonLink fucked around with this message at 17:18 on Mar 10, 2023

abelwingnut
Dec 23, 2002


isn’t beating that go master one of ai’s biggest achievements so far? not sure how it did that, but pretty wild from my novice perspective.

Ghost Leviathan
Mar 2, 2017
Probation
Can't post for 18 hours!
That's just that old saying about how great swordsmen worry about amateurs, because they have no idea what that idiot is going to do.

Watermelon Daiquiri
Jul 10, 2010
I TRIED TO BAIT THE TXPOL THREAD WITH THE WORLD'S WORST POSSIBLE TAKE AND ALL I GOT WAS THIS STUPID AVATAR.

Family Values posted:

Yes... but no. People play against other people in strategy games (thinking Paradox-style multiplayer grand strategy games) if they want a real challenge. The problem is that competent competitive play exposes how unbalanced those games usually are and that there's an optimal, not very fun or 'flavorful' way to win and anything else is a loser's strategy (like EU4 multiplayer you either go quantity or go home).

An actually challenging AI would require similar unfun play in single player games.

A way around it might be to have multiple AI personalities. Some that were trained for varying amounts of time, some that were trained in restrictive environments (only allowed to take certain choices and never others), and a few that are absolute beasts. Then which AI you get will determine your style of play.

What about people who want to face human opponents but with the ability to pause the game for an indefinite amount of time and pick it back up whenever?

Arsenic Lupin
Apr 12, 2012

This particularly rapid💨 unintelligible 😖patter💁 isn't generally heard🧏‍♂️, and if it is🤔, it doesn't matter💁.



Too late; bank went boom.
https://twitter.com/BrendanPedersen/status/1634236377066881024
e: holy poo poo
https://twitter.com/tommyrulznyc/status/1634238029903585280

Arsenic Lupin fucked around with this message at 18:27 on Mar 10, 2023

Rent-A-Cop
Oct 15, 2004

I posted my food for USPOL Thanksgiving!

I see mortgage backed securities. Did this bank just do a speedrun of 2008?

Also on the game AI chat the issue I've always had especially in the bigger strategic games is that no matter how dumb the AI is if you let it get big enough it will just out micro you with a million stupid moves a minute.

StumblyWumbly
Sep 12, 2007

Batmanticore!
Who would have thought that the independent free thinkers at Silicon Valley would all pile into one bank and then suddenly run out.

Gort
Aug 18, 2003

Good day what ho cup of tea

Rent-A-Cop posted:

Also on the game AI chat the issue I've always had especially in the bigger strategic games is that no matter how dumb the AI is if you let it get big enough it will just out micro you with a million stupid moves a minute.

That reminds me of the famous Gary Kasparov vs Deep Blue chess matches where the sheer speed of Deep Blue's moves completely threw off Kasparov, who was used to having both his own thinking time but also the opponent's thinking time to plan his moves. Deep Blue came back with its moves so quickly he got a fraction of the time to think that he usually did.

Player attention is a resource, and an AI can effectively have infinite amounts of it.

Cabbit
Jul 19, 2001

Is that everything you have?


I wonder if some big tech executives just got a huge chunk of their wealth wiped out.

OddObserver
Apr 3, 2009

Rent-A-Cop posted:

I see mortgage backed securities. Did this bank just do a speedrun of 2008?



From what I understood, not really, in that those securities don't have any sort of problem, but they are long-term commitments which is kinda an issue for a bank that may have people come in and ask for X bucks to pay the bills now.

Aramis
Sep 22, 2009



OddObserver posted:

From what I understood, not really, in that those securities don't have any sort of problem, but they are long-term commitments which is kinda an issue for a bank that may have people come in and ask for X bucks to pay the bills now.

Yes and no. A long term loan at a lower interest rate than the Fed rate might as well be almost worthless until the due date since no one in their right mind would want to buy them unless they get them at a huge discount.

Aramis fucked around with this message at 19:59 on Mar 10, 2023

Arsenic Lupin
Apr 12, 2012

This particularly rapid💨 unintelligible 😖patter💁 isn't generally heard🧏‍♂️, and if it is🤔, it doesn't matter💁.


Cabbit posted:

I wonder if some big tech executives just got a huge chunk of their wealth wiped out.
The news coverage I've seen says that CEOs often used it as their money manager, their mortgage lender, as well as for their startup. It also says that it was sometimes a requirement of angel financing that you bank at SVB.

cat botherer
Jan 6, 2022

I am interested in most phases of data processing.

Arsenic Lupin posted:

The news coverage I've seen says that CEOs often used it as their money manager, their mortgage lender, as well as for their startup. It also says that it was sometimes a requirement of angel financing that you bank at SVB.
What made SVB better than any other loving bank they could park their money?

Fate Accomplice
Nov 30, 2006




Arsenic Lupin posted:

The news coverage I've seen says that CEOs often used it as their money manager, their mortgage lender, as well as for their startup. It also says that it was sometimes a requirement of angel financing that you bank at SVB.

https://twitter.com/88888sAccount/status/1634028258500169731

OddObserver
Apr 3, 2009

Aramis posted:

Yes and no. A long term loan at a lower interest rate than the Fed rate might as well be almost worthless until the due date since no one in their right mind would want to buy them unless unless they get them at a huge discount.

Aha, thanks, I was missing the connection to interest rates. So if I understood your clarification it went something like this:
1) Bank invests in long-term stuff that has low interest rate return but is low risk. The expectation is that since it's low risk it would be easy to sell to someone at nominal price if they need cash to meet consumer demands.
2) Fed raises rates
3) Now it's hard to sell the debt since potential buyers can get better returns elsewhere (US bonds?)

... It feels like something the FDIC involvement could theoretically fix then, in that holding on to stuff for long-term is in principle OK for US government (but I don't know if that's something that's actually permitted absent an explicit congressional bailout).

Also I am right to say that doing #1 while interest rates were super-low was a dubious decision? Though I feel like normally you want banks doing low risk/low return stuff, so what am I missing here?

Aramis
Sep 22, 2009



cat botherer posted:

What made SVB better than any other loving bank they could park their money?

My understanding is that a lot of VCs were also invested in the bank itself. So a lot of funding was conditional on it so that the VC ends up paying itself.


You got the gist of it.

OddObserver posted:

... It feels like something the FDIC involvement could theoretically fix then, in that holding on to stuff for long-term is in principle OK for US government (but I don't know if that's something that's actually permitted absent an explicit congressional bailout).

In principle, but that would require the federal reserve to get involved since the FDIC can't print new money.

OddObserver posted:

Also I am right to say that doing #1 while interest rates were super-low was a dubious decision? Though I feel like normally you want banks doing low risk/low return stuff, so what am I missing here?

Kinda sorta. Banks have to give "safe" loans, but short term safe loans when interest rates are low have near worthless returns. They would have been fine if there wasn't a run on the bank, which not a single bank is immune to.

Aramis fucked around with this message at 20:02 on Mar 10, 2023

Magic Underwear
May 14, 2003


Young Orc

Arsenic Lupin posted:

The news coverage I've seen says that CEOs often used it as their money manager, their mortgage lender, as well as for their startup. It also says that it was sometimes a requirement of angel financing that you bank at SVB.

Are we going to see shittech contagion from this? What's going to happen to my subscriptions to Netflix but for dogs and uber but for vitamin supplements?

GhostofJohnMuir
Aug 14, 2014

anime is not good

OddObserver posted:

Aha, thanks, I was missing the connection to interest rates. So if I understood your clarification it went something like this:
1) Bank invests in long-term stuff that has low interest rate return but is low risk. The expectation is that since it's low risk it would be easy to sell to someone at nominal price if they need cash to meet consumer demands.
2) Fed raises rates
3) Now it's hard to sell the debt since potential buyers can get better returns elsewhere (US bonds?)

... It feels like something the FDIC involvement could theoretically fix then, in that holding on to stuff for long-term is in principle OK for US government (but I don't know if that's something that's actually permitted absent an explicit congressional bailout).

Also I am right to say that doing #1 while interest rates were super-low was a dubious decision? Though I feel like normally you want banks doing low risk/low return stuff, so what am I missing here?

a lot of the problem is that svb had all of its depositors concentrated in very specific niche. it made it much more likely that many of their deposits would move in the same direction at once. historically startups have done poorly in a rising rate environment, and long term treasuries have done poorly in a rising rate environment. not my area of expertise, but it suggests that they either should have sought out more diversified depositors, or put a portion of their assets into something that hedges against a rising interest rates

Arsenic Lupin
Apr 12, 2012

This particularly rapid💨 unintelligible 😖patter💁 isn't generally heard🧏‍♂️, and if it is🤔, it doesn't matter💁.


Magic Underwear posted:

Are we going to see shittech contagion from this? What's going to happen to my subscriptions to Netflix but for dogs and uber but for vitamin supplements?

Some of those companies won't make payroll Monday. This is a highly technical thread; Rippling is one of the providers for payroll services, and they used SVB as their bank.

https://twitter.com/parkerconrad/status/1634237386564730882

The most significant tweet:
https://twitter.com/parkerconrad/status/1634238256937308160

"In flight" means "we initiated the bank transfer, but it hadn't yet shown up at the destination". A lot of paychecks won't show up.

https://twitter.com/parkerconrad/status/1634240849147219979

e: And in a different nightmare:
https://twitter.com/KimZetter/status/1634300361589530626

Arsenic Lupin fucked around with this message at 22:12 on Mar 10, 2023

Tacier
Jul 22, 2003

So because the money tied up in those securities isn’t really gone, but rather just impossible to sell and turn into immediate liquidity, does that mean investors could eventually be made whole again? Who would administer those payouts if the bank no longer exists? Or am I misunderstanding?

evilweasel
Aug 24, 2002

Tacier posted:

So because the money tied up in those securities isn’t really gone, but rather just impossible to sell and turn into immediate liquidity, does that mean investors could eventually be made whole again? Who would administer those payouts if the bank no longer exists? Or am I misunderstanding?

a bank of last resort - such as the fed, or a larger bank the FDIC agrees to underwrite - could profitably take on the assets and make depositors whole. this is because the loans aren't bad, they're just not as profitable as lending money to the government at current interest rates.

Boris Galerkin
Dec 17, 2011

I don't understand why I can't harass people online. Seriously, somebody please explain why I shouldn't be allowed to stalk others on social media!
Can someone explain to someone stupid about banking and investments why this bank failed and who’s gonna get hosed over? Like if it’s average Joe that gets screwed over I’m not gonna laugh but if it’s startup founders (who lets be real are only going to be mildly inconvenienced) then I wanna drink some schadenfreude tea.

Fate Accomplice
Nov 30, 2006




Boris Galerkin posted:

Can someone explain to someone stupid about banking and investments why this bank failed and who’s gonna get hosed over? Like if it’s average Joe that gets screwed over I’m not gonna laugh but if it’s startup founders (who lets be real are only going to be mildly inconvenienced) then I wanna drink some schadenfreude tea.

https://archive.ph/Ger2q#selection-4001.0-4005.852

from today's excellent Levine column: SVB is called "the bank of Startups" here cause that's what they focused on serving.

levine posted:

Crudely stereotyping, in traditional banking, you take deposits and make loans. In the Bank of Startups, in 2021, you take deposits and mostly buy bonds. Again crudely stereotyping, corporate loans often have floating interest rates and shorter terms, while bonds have fixed interest rates and longer terms. None of this is completely true — there are fixed-rate corporate loans and floating-rate bonds, traditional banking tends to involve making lots of loans (like mortgages) with long-term fixed rates, you can do swaps, etc. — but it is a useful crude stereotype. 3

Or, to put it in different crude terms, in traditional banking, you make your money in part by taking credit risk: You get to know your customers, you try to get good at knowing which of them will be able to pay back loans, and then you make loans to those good customers. In the Bank of Startups, in 2021, you couldn’t really make money by taking credit risk: Your customers just didn’t need enough credit to give you the credit risk that you needed to make money on all those deposits. So you had to make your money by taking interest-rate risk: Instead of making loans to risky corporate borrowers, you bought long-term bonds backed by the US government.

The result of this is that, as the Bank of Startups, you were unusually exposed to interest-rate risk. Most banks, when interest rates go up, have to pay more interest on deposits, but get paid more interest on their loans, and end up profiting from rising interest rates. But you, as the Bank of Startups, own a lot of long-duration bonds, and their market value goes down as rates go up. Every bank has some mix of this — every bank borrows short to lend long; that’s what banking is — but many banks end up a bit more balanced than the Bank of Startups.

But there is another, subtler, more dangerous exposure to interest rates: You are the Bank of Startups, and startups are a low-interest-rate phenomenon. When interest rates are low everywhere, a dollar in 20 years is about as good as a dollar today, so a startup whose business model is “we will lose money for a decade building artificial intelligence, and then rake in lots of money in the far future” sounds pretty good. When interest rates are higher, a dollar today is better than a dollar tomorrow, so investors want cash flows. When interest rates were low for a long time, and suddenly become high, all the money that was rushing to your customers is suddenly cut off. Your clients who were “obtaining liquidity through liquidity events, such as IPOs, secondary offerings, SPAC fundraising, venture capital investments, acquisitions and other fundraising activities” stop doing that. Your customers keep taking money out of the bank to pay rent and salaries, but they stop depositing new money.

essentially:
1 SVB takes tons of money from VCs and startups getting money from VCs
2 those entities don't really borrow more money in the way banks want to lend it
3 SVB makes its return by investing in interest rate risk
4 interest rates rise
5 oh poo poo our customers (VCs and startups) only can exist in low interest environments
6 so not only are the value of our interest rate investments going down, our customers want their money back
7 we are here

Fate Accomplice fucked around with this message at 23:05 on Mar 10, 2023

Aramis
Sep 22, 2009



Bank put a lot of their customer's money in metaphorical vaults with timed locks, but too many customers want to get their money NOW. In principle they could get their money eventually but:

1) We are talking years, which a lot of the customers can't afford to wait.,
2) The bank had to sell some of these vaults for less money than is inside of them (60 cents on the dollar) to fulfill withdrawal requests, so they were actually running out of cash, which is why they got shut down in the middle of a day.

The main consequence is that a bunch of companies are going to need to borrow a bunch of money right away if they want to be able to pay their bills, which means taking a bunch of unfavourable debt. Depending on how tight their finances are, they may or may not be able to weather this,.

It will mean absolutely nothing for regular people unless employed by these companies.

Riven
Apr 22, 2002
Yeah after reading this I immediately DM'd my CEO asking if we bank with them. Thankfully neither we nor our payroll provider (Gusto) do.

Twerk from Home
Jan 17, 2009

This avatar brought to you by the 'save our dead gay forums' foundation.

Tacier posted:

So because the money tied up in those securities isn’t really gone, but rather just impossible to sell and turn into immediate liquidity, does that mean investors could eventually be made whole again? Who would administer those payouts if the bank no longer exists? Or am I misunderstanding?

The bank used deposits to buy a ton of long term mortgage backed securities that are paying like 1.5% interest.

So, yeah, depositors will get their money back, in 15 years when it's been devastated by inflation.

cat botherer
Jan 6, 2022

I am interested in most phases of data processing.

Twerk from Home posted:

The bank used deposits to buy a ton of long term mortgage backed securities that are paying like 1.5% interest.

So, yeah, depositors will get their money back, in 15 years when it's been devastated by inflation.
They'll get (most of) the money back before that. The long duration assets can still be sold off for cash, it's not like you have to hold them to maturity.

Motronic
Nov 6, 2009

cat botherer posted:

They'll get (most of) the money back before that. The long duration assets can still be sold off for cash, it's not like you have to hold them to maturity.

But they can't be sold for the amount of money they need..........that's the rub.

cat botherer
Jan 6, 2022

I am interested in most phases of data processing.

Motronic posted:

But they can't be sold for the amount of money they need..........that's the rub.
Yeah but getting the bulk of the value back won't take 15 years.

e: Here is deepfake Tucker Carlson to tell you which Pokemon is the best to gently caress. Or maybe it's the real Tucker, who can really say? (nsfw obv)

https://www.youtube.com/watch?v=DynOlXtlYTs

Motronic
Nov 6, 2009

cat botherer posted:

Yeah but getting the bulk of the value back won't take 15 years.

....and? So what. You still have to take the haircut and as a bank when you can't pay your bills with what they were sold for someone else has to.

Which historically ends up as some larger bank getting a gift and the taxpayers footing the rest.

cat botherer
Jan 6, 2022

I am interested in most phases of data processing.

Motronic posted:

....and? So what. You still have to take the haircut and when you can't pay your bills as a bank someone else has to.

Which historically ends up as some larger bank getting a gift and the taxpayers footing the rest.
The post I was responding to literally said it would take 15 years. We're not in disagreement here unless you think it will take 15 years.

Motronic
Nov 6, 2009

cat botherer posted:

The post I was responding to literally said it would take 15 years. We're not in disagreement here unless you think it will take 15 years.

It's going to for someone, and someone else has to pay for that to happen.

Maybe I'm missing your point, or maybe you're not understanding that this can't just "go away" because you can sell things at a loss.

In the end, someone has to pay for this taking 15 years and someone (probably someone else who has been paid or several derivatives thereof) has to wait 15 years.

Motronic fucked around with this message at 00:58 on Mar 11, 2023

cat botherer
Jan 6, 2022

I am interested in most phases of data processing.

Motronic posted:

It's going to for someone, and someone else has to pay for that to happen.

Maybe I'm missing your point, or maybe you're not understanding that this can't just "go away" because you can sell things at a loss.

In the end, someone has to pay for this taking 15 years.
The long-maturing assets take that long, but they'll be sold off to entities who want that in their portfolio. Some assets take a long time to mature, but that's appropriate in some situations, which is why those kind of assets exist.

You're really putting words in my mouth here. I never said this would "go away." I've got no idea where you got that from.

cat botherer posted:

They'll get (most of) the money back before that. The long duration assets can still be sold off for cash, it's not like you have to hold them to maturity.
Notice how I said (most of)? You might want to try reading the context here.

Adbot
ADBOT LOVES YOU

Aramis
Sep 22, 2009



cat botherer posted:

The long-maturing assets take that long, but they'll be sold off to entities who want that in their portfolio. Some assets take a long time to mature, but that's appropriate in some situations, which is why those kind of assets exist.

Sell these underwater assets to who? Aquaman?

I know they are not technically underwater, but the joke was right there. However, my point stands: who would want those apart from at a massive discount?

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply