@gimulnautti@mastodon.green
@gimulnautti@mastodon.green avatar

gimulnautti

@gimulnautti@mastodon.green

Pirate Politician, Software Developer, MA Musicology (he/him)

Interests:
Culture, Math, Sociology, Philosophy, Music, Demoscene, Games, #NAFO, Writing, Disc Golf, Skateboarding.

Not here for snarky comebacks. A slur is not an argument.

Other fediverse accounts:
https://mastodon.green/@gimulnautti @metapixl.com
https://mastodon.green/@gimulnautti @mastodontti.fi

Blog:
https://gimulnaut.wordpress.com
https://medium.com/@toni.k.aittoniemi

This profile is from a federated server and may be incomplete. Browse more on the original instance.

gimulnautti, to programming
@gimulnautti@mastodon.green avatar

As a programmer, I use -based autocomplete daily.

My career spans almost 30 years now, and there is a issue in which is usually referred to as . It is a combination term from A) code which means the code has bad organisation and lacks structural rules B) mindlessly copying something.

Enough ”doing the same thing over and over again” can slowly turn even the best codebase into a badly organised mess, as the code no longer reflects the big picture.

gimulnautti,
@gimulnautti@mastodon.green avatar

that I use is simply perfect for creating copypasta. This can probably boost efficiency for a some junior-level tasks, but relying on it too much can also risk deteriorating the quality of your codebase.

Extrapolating from this to what our society is going to face as ML tools augment our communication, I feel we might need to be prepared for a lot more copypasta that does the job, but with itself in ways that make it harder to see what is actually happening.

davidak, to ai
@davidak@chaos.social avatar

is Smarter than You Think: Introducing

https://www.youtube.com/watch?v=wVzuvf9D9BU

gimulnautti,
@gimulnautti@mastodon.green avatar

@davidak Thinking in prompt systems is something I’ve been also trying to explain to people lately. Thanks for this one, it’s almost the perfect example to show to them!

alsutton, to ai
@alsutton@snapp.social avatar

Anyone else betting models this decade will follow the trajectory of social media last decade; In 10 years time folk will be getting what the tech-savvy folk wanted from the start - an open solution everyone can contribute to and use as they want?

gimulnautti,
@gimulnautti@mastodon.green avatar

@Sh41 @alsutton yes, the recent google leak speculated 93% of gpt-4 level capability of the best open-source models.

gimulnautti, to random
@gimulnautti@mastodon.green avatar

Within a year or two, we will be able to actually talk to animals. We won’t be able to really understand what they mean, but the language probably won’t be the problem.

https://open.spotify.com/episode/2tCoKxrVVCUYhfi575XUID?si=cLlQOLarTXW0icia4F21Qw

gimulnautti,
@gimulnautti@mastodon.green avatar

@JohanEmpa Yes. But the model input would be more than that. Think:

Dog A is the real dog
Dog B is the ML

You specify the position, speed, posture and other aspects of dog B, and the model will spit out how the dog would move and bark in the next 2 seconds.

Diggler67, to random
@Diggler67@thecanadian.social avatar

This video by Tristan Harris and Aza Raskin helped me understand why the current "arms race" in implementation is so dangerous. 😬

https://youtu.be/xoVJKj8lcNQ

I'm probably late to the party on this, but I just saw it linked somewhere today (lost the reference) and just watched it tonight. These guys really know how to explain things to non-technical people. It's really worth watching, but you need to set aside an hour!

I had a vague layman's uneasiness about until now...

gimulnautti,
@gimulnautti@mastodon.green avatar

@Diggler67 @panamared27401 This is legit one of the best popularisations of the field of problems underlying the worries.

I like how it really underlines that we are largely looking at the wrong moment in AI: where it becomes better than human excellence.

The moment that counts though, has maybe already passed; Where it surpasses human weaknesses.

drahardja, to photography
@drahardja@sfba.social avatar

This article describes how journalists sent photographs across the Atlantic “instantly” in 1926. The most amazing thing to me is that the image is literally digitized into “pixels” along a helical scanline, quantized into 5 levels of gray—all through analog means.

So utterly fascinating.

An enterprising person could have used the 5 hole punches to encode 32 levels of gray, or used maybe 3 holes to encode 8 levels of gray, saving 40% of the transmission cost at the expense of making the machines more complicated…

https://petapixel.com/2013/07/01/how-they-sent-photos-across-the-ocean-back-in-1926/

gimulnautti,
@gimulnautti@mastodon.green avatar

@drahardja @retrogaming_bot super cool!

estherschindler, to random
@estherschindler@hachyderm.io avatar

Yes, children, we had to pay extra to get sound on our computers.

gimulnautti,
@gimulnautti@mastodon.green avatar

@estherschindler @damonyoung Unless you had an

gimulnautti, to opensource
@gimulnautti@mastodon.green avatar

With the fast advances in low-cost #opensource #training techniques for #machinelearning, its apparent that #alignment will not be solved by gate-keepers on what goes into #RLHF and #compute.

Verifying humans and truthfullness of reporting should be our no.1 priority.

https://simonwillison.net/2023/May/4/no-moat/

gimulnautti, to random Finnish
@gimulnautti@mastodon.green avatar
valoisa, to random
@valoisa@eliitin-some.fi avatar

Populisteilla on meneillään paha paikka kun pitäisi samaan aikaan esittää vahvaa huutamalla pää punaisena ja yrittää saada ulkoa syy päästä oppositioon. Hehän eivät koskaan oikeasti halua demokraattiseen hallitukseen, koska se tarkoittaa kompromisseja ja heidän äänestäjänsä eivät osaa tehdä kompromisseja.

gimulnautti,
@gimulnautti@mastodon.green avatar

@valoisa Ja sitten Riikka tanssii Mussolinin.

gimulnautti,
@gimulnautti@mastodon.green avatar

@valoisa Ei se koskaan ole ollut pihalla. Perussuomalaisen puolueen tavoite ja linja ovat olleet selviä alusta alkaen: Poistaa muilta kuin itseltään.

hanse_mina, to random
@hanse_mina@nafo.uk avatar

RT WarFrontline: 🚨, "After we took over a Russian trench, the Belorussian commander used a radio he found and pretended to be Russian and gave false coordinates to the Russian artillery. It worked, they knocked out another Russian unit.",
-Captain Pavel Szurmiej‼️🍻😍

gimulnautti,
@gimulnautti@mastodon.green avatar

@hanse_mina I think there are still many out there in the Russian lines who would rather see Ukraine win.

atrupar, to random
@atrupar@journa.host avatar

Fox News's advice for life in America: "Have a plan to kill everyone you meet"

video/mp4

gimulnautti,
@gimulnautti@mastodon.green avatar

@atrupar @juuhaa Full circle.

OnceUponAGoblin, to random Portuguese
@OnceUponAGoblin@masto.pt avatar

I never get tired of this photo. Steve Jobs showing off the first Macintosh to Andy Warhol, Sean Lennon, and Keith Haring in 1984.

gimulnautti,
@gimulnautti@mastodon.green avatar

@OnceUponAGoblin @retrogaming_bot I bet someone at that age shouted: ”Get a real job!”

artomelaranta, to random Finnish

Siinä vaiheessa kun kansanedustaja huutaa kovaan ääneen että poliitikkojen tulisi puuttua jonkin median lähettämään sisältöön ollaan jo koputtelemassa fasismin ovella

gimulnautti,
@gimulnautti@mastodon.green avatar

@artomelaranta @juuhaa Ja sitten Riikka tanssii Mussolinin.

JamesGleick, to random
@JamesGleick@zirk.us avatar

Once again Ted Chiang has it exactly right. The immediate danger from is not that it will become sentient and do whatever it wants. The danger is that it will do what it’s being designed to do: help rich corporations destroy the working class in pursuit of ever-greater profits and thus concentrate wealth in fewer and fewer hands.

https://www.newyorker.com/science/annals-of-artificial-intelligence/will-ai-become-the-new-mckinsey

gimulnautti,
@gimulnautti@mastodon.green avatar

@JamesGleick with the recent advances in , AI will be available to everyone. And if I know human nature at all it will be first applied by malicious actors, only one of them being capital as specified here.

The steps taken by society to mitigate should be of humans and verification of accurate reporting.

AI for every kind of social and political mischief imaginable will be available by 2025.

Governments need to act, stop looking at business.

ellent, to random Dutch

The UK and US have intervened in the race to develop ever more powerful technology, as the British competition watchdog launched a review of the sector and the White House advised tech firms of their fundamental responsibility to develop safe products

https://www.theguardian.com/technology/2023/may/04/uk-and-us-intervene-amid-ai-industrys-rapid-advances

gimulnautti,
@gimulnautti@mastodon.green avatar

@ellent Safety protocols for commercial actors in the AI space are one aspect of the problem, but as a whole just regulating companies isn’t even cutting it close.

The community has answered the call of developing lighter techiques, which I predict will make training AI capable of fooling 98% of the population available to anyone by 2025.

Yes, this means the nasty people. Populists, fascists, nazis.

We need systems for humans and truthfull reporting, now.

gimulnautti, to random
@gimulnautti@mastodon.green avatar

@tootapp I want to block a user that is sending me pm spam. Do I have to do it in the webapp? The client doesn’t seem to allow it?

gimulnautti,
@gimulnautti@mastodon.green avatar

@tootapp I don’t have it for the spammer.

gimulnautti,
@gimulnautti@mastodon.green avatar

@tootapp seems to be the case. Thank you and thanks for the admins!

asahi95, to fediverse

Attention , please be on the lookout for fake crypto scam bot accounts coming from mastodon.social

Remember to check the username and make sure to report and block these account as much as you can!

:BoostOK: Be sure to boost this for more attention

gimulnautti,
@gimulnautti@mastodon.green avatar

@asahi95 yup. Got one.

StillIRise1963, to random
@StillIRise1963@mastodon.world avatar

John Roberts believes those entrusted with the law are above the law.

gimulnautti,
@gimulnautti@mastodon.green avatar

@StillIRise1963 He what?

I mean Montesquieu came up with the basics framework of the modern state in the mid 1700’s.

The enforcement of law or interpretation of judgement can in no way be tied to regulators or the branches tied to each other.

Autocracy happens when principle of rule of law that’s created by the separation of branches breaks down, basically.

Can we afford to forget things like this? I think not.

gimulnautti, to science
@gimulnautti@mastodon.green avatar

I nice reminder, that since late most papers in advanced AI have come from people working at: Mega large companies for profit.

Which raises an important question: How is profit and science supposed mix, if peer-review has been impossible due to witholding data?

If a paper can’t be verified, is it science? Science is, after all, an inter-subjective process.

So are these papers more truthfully just fancy forms of advertisements?

https://www.vice.com/en/article/wxjdg5/scary-emergent-ai-abilities-are-just-a-mirage-produced-by-researchers-stanford-study-says

gimulnautti,
@gimulnautti@mastodon.green avatar

@Dinkenfunkle Yes. What I am emphasizing is that in regards to LLM’s we are very much in the research & engineering stage.

For this particular instance however, the progress is inverted from the usual. Science seems like it’s going to happen after this time.

In general people tend to not grasp that science means the communal effort of independent verification, so I thought it important to emphasize the difference.

vs

mnutty, to politics

1/ Today the kicks off in earnest with ’s re-election bid announcement

There is 560 days to the . Much will happen, conditions will change but it as suggests a continuation of the battle for the soul of America

I’ll use this thread to comment periodically on the election over the next 80 weeks. It’s a long road to , but the stakes couldn’t be higher

https://youtu.be/ChjibtX0UzU

gimulnautti,
@gimulnautti@mastodon.green avatar

@volkris @mnutty Signing and verifying human-generated content should have been top-priority last year already.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • JUstTest
  • thenastyranch
  • magazineikmin
  • ethstaker
  • khanakhh
  • rosin
  • Youngstown
  • everett
  • slotface
  • ngwrru68w68
  • mdbf
  • GTA5RPClips
  • kavyap
  • DreamBathrooms
  • provamag3
  • cisconetworking
  • cubers
  • Leos
  • InstantRegret
  • Durango
  • tacticalgear
  • tester
  • osvaldo12
  • normalnudes
  • anitta
  • modclub
  • megavids
  • lostlight
  • All magazines
    •