dcl

As an ML person I totally get the how NN works to accomplish this and it's very cool. What's really cool is how they get this to work in real time with little/no latency.

Only a small matter of time now before we have Vec2TikTok (or similar) where you'll be served up videos specifically optimised with the kind of content that will keep you scrolling or watching more.

People will end up watching news about events that never happened or product reviews for things that don't exist.

I've worked at a bank. A significant proportion of the the workforce had no idea how it works. I'd say <10% of the senior executives have the slightest clue.

But it's not 'as they wish' as you say, depending on where they operate, they are typically constrained by capital and liquidity ratios set by their regulator.

According to the wiki article on multi-armed bandits, apparently the Allies contemplated this in WW2. https://en.wikipedia.org/wiki/Multi-armed_bandit#Empirical_m...

> Originally considered by Allied scientists in World War II, it proved so intractable that, according to Peter Whittle, the problem was proposed to be dropped over Germany so that German scientists could also waste their time on it.[13]

[13] Whittle, Peter (1979), "Discussion of Dr Gittins' paper", Journal of the Royal Statistical Society, Series B, 41 (2): 148–177, doi:10.1111/j.2517-6161.1979.tb01069.x

This paper discusses that exact issue:

Insurance makes wealth grow faster Ole Peters, Alexander Adamou https://arxiv.org/abs/1507.04655

Yep one of those cases, where an ensemble average is not at all relevant for describing the situation.

Chris Dixon (https://twitter.com/cdixon) posts some particularly outrageous and as far as I'm concerned, disingenuous comments on this topic. I would be ashamed to be associated with him.

Indeed, I believe/think TikTok is probably the most psychologically harmful social media app.

Obviously none allow it. But it's silly not fear the logical conclusion of the CCP's actions.

True. But I fear China more.

Chilling, and I suspect that western governments will do very little to prevent this kind of extra-territorial control exerted by the CCP.

In a few years I expect it will be very increasingly difficult to criticise China.

I remember that blog from when I was studying penalty/regularization methods over a decade ago. Was a great source of interesting papers and ideas.

Indeed. I'd like to think that with very good constraints on things like aiming and timing, you might be able to help foster creative 'gameplay' and discovery of things like map routes, map control etc.

I vaguely recall this now. Thanks for the reminder.

I'd love to see how far/good bots could get in a deathmatch shooter with reinforcement learning/self play (provided we could strongly and intelligently enforce human level reaction times and keyboards/mouse inputs). Would be really interesting to see if any crazy new strategies/tactics emerge.

Do you have any favourite libraries for producing such?

This echoes my experience. It often finds seasonality that doesn't exist, but it also makes it look like you are doing something impressive...

Inexperienced stakeholders don't want/like to see smooth forecasts, even if you provide predication intervals.

Utterly disgusting.

As far as I'm aware, the Australian government cannot do much about this either (not that they would chose to). There are laws about not interfering to preserve academic freedom (kinds of ironic here).

This is made worse by our ex Prime Minister refusing to admit that this is a serious issue. (He is close personal friends with at least one University head). https://podcasts.apple.com/au/podcast/86-coronavirus-covert-...

It's not just the UI that's inconsistent, it's that the different UI layers can actually present contradictory information.

https://i.imgur.com/oNxlUzv.png

Outside of forcing a stocksplit, what do people anticipate an architectural solution would be? Is it simply having to rewrite a bunch of code to deal with 64bit numbers?

maybe try screening for people that are less likely to be affected by it first?

I say try because it probably isn't going to be easy.

HSBC then.

> HSBC Holdings Plc agreed to pay a record $1.92 billion in fines to U.S. authorities for allowing itself to be used to launder a river of drug money flowing out of Mexico and other banking lapses.

https://www.reuters.com/article/us-hsbc-probe-idUSBRE8BA05M2...

I see. So what would you even really own once that URL goes down.

The whole thing relies on a weird level of trust and belief in what other people believe to be valuable.

I really don't understand this. What mechanism is there to ensure an NFT was created by the 'right person/people'? (whatever this means).

I could make an NFT for Mickey Mouse, and it _should_ be worthless, but if Disney created one then maybe it could be worth something?

I've got some dumb questions about this whole NFT thing...

Is the artwork actually stored in the token or does it contain a URL that isn't guaranteed to always exist? I'm guessing there is no mechanism to change the link if the host should go down or stop hosting the content.

Are NFT's used to recognise a transfer of ownership of the intelectual property?

Neat. I've noticed that a few stock market apps/sites definitely use different methods to do this, and with dramatically different results.

Indeed. I'll often see experts in their fields and otherwise very interesting people retweet and reply to stuff I absolutely don't care about, or have already heard enough of.

I've found the algo generated list of tweets that I'm presented with in the app to be pretty poor these days. More and more often I am having to manually go to specific users to find their tweets because they are not being uplifted by the algorithms.

Neat idea. A real good test of NLP to identify similar posts/tweet and posters. Signals could be generating by swiping left or right based on interest of the post, or clicking on a star or heart to indicate 'super interest'.

Will no doubt end up in situations where people are complaining about _other peoples_ filter bubbles though.

Wow, you're not wrong.