Lmst

What are some good #AI accounts to follow? Especially #watchdog monitoring #AGI and #Superintelligence development? #AI2027

I think that one of the reasons that I’m not worried about AGI is that I’m still waiting for the end of the world to be caused by Quantum Computing
https://alecmuffett.com/article/113401
#ItsTheEndOfTheWorldAsWeKnowItAndIFeelFine #QuantumComputing #ai2027

I think that one of the reasons that I’m not worried about AGI is that I’m still waiting for the end of the world to be caused by Quantum Computing

The apocalypse is a perennial favorite for pulp authors:

https://ai-2027.com/

#ai2027 #feed #itSTheEndOfTheWorldAsWeKnowItAndIFeelFine #quantumComputing

The "AI 2027" Scenario: How realistic is it?

https://garymarcus.substack.com/p/the-ai-2027-scenario-how-realistic

#HackerNews #AI2027 #AIrealism #FutureTech #TechnologyTrends #Innovation

also omfg this read is really terrifying, or amazing depending on which side you lean, and are actually written by people that know what they are talking about:

https://ai-2027.com/

#tech #llm #ai #openai #gpt #chatgpt #future #prediction #2027 #ai2027

@bojerlanski
Ich kann dazu diesen sehr ausführlichen und interessanten Report empfehlen:
iceshrimp.de/notes/a7m32xdsz0zx18ce

#AI2027

AI 2027 - Intro:
"We predict that the impact of superhuman AI over the next decade will be enormous, exceeding that of the Industrial Revolution.
We wrote a scenario that represents our best guess about what that might look like. It’s informed by trend extrapolations, wargames, expert feedback, experience at OpenAI, and previous forecasting successes."

by Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, Romeo Dean

---

"Late 2025: The World’s Most Expensive AI
OpenBrain is building the biggest datacenters the world has ever seen. (To avoid singling out any one existing company, we’re going to describe a fictional artificial general intelligence company, which we’ll call OpenBrain. We imagine the others to be 3–9 months behind OpenBrain.)
GPT-3 (3 x 10^23 FLOPS)
GPT-4 (2 x 10^25 FLOPS)
Agent-1 (3 x 10^27 FLOPS)

GPT-4 required 2⋅1025 FLOP of compute to train. OpenBrain’s latest public model—Agent-0—was trained with 1027 FLOP."
(read more in the description of figure 1)

---

"In our timelines forecast, we predict when OpenBrain will internally develop a superhuman coder (SC): an AI system that can do any coding tasks that the best AGI company engineer does, while being much faster and cheaper."
(read more in the description of figure 2)

---

"Aided by the new capabilities breakthroughs, Agent-3 is a fast and cheap superhuman coder. OpenBrain runs 200,000 Agent-3 copies in parallel, creating a workforce equivalent to 50,000 copies of the best human coder sped up by 30x.
(...)
Still, they are achieving a year’s worth of algorithmic progress every week and will therefore soon be up against the limits of the Agent-4 paradigm.
(...)
As Agent-4 gets smarter, it becomes harder for Agent-3 to oversee it. For example, Agent-4’s neuralese “language” becomes as alien and incomprehensible to Agent-3 as Agent-3’s is to humans.
(...)
Mar 2027: Superhuman coder (SC):
An AI system that can do the job of the best human coder on tasks involved in AI research but faster, and cheaply enough to run lots of copies.

Aug 2027: Superhuman AI researcher (SAR):
The same as SC but for all cognitive AI research tasks.

Nov 2027: Superintelligent AI researcher (SIAR):
An AI system that is vastly better than the best human researcher at AI research.

Dec 2027: Artificial superintelligence (ASI): An AI system that is much better than the best human at every cognitive task.
(...)
Things that sound like science fiction keep happening in real life."

#AI2027 #KünstlicheIntelligenz #KI #ArtificialIntelligence #SuperhumanCoder #SuperhumanAIresearcher #SuperintelligentAIresearcher #ArtificialSuperintelligence

Late 2025: The World’s Most Expensive AI
OpenBrain is building the biggest datacenters the world has ever seen. (To avoid singling out any one existing company, we’re going to describe a fictional artificial general intelligence company, which we’ll call OpenBrain. We imagine the others to be 3–9 months behind OpenBrain.)
GPT-3 (3 x 10^23 FLOPS)
GPT-4 (2 x 10^25 FLOPS)
Agent-1 (3 x 10^27 FLOPS)

GPT-4 required 2⋅1025 FLOP of compute to train. OpenBrain’s latest public model—Agent-0—was trained with 1027 FLOP. Once the new datacenters are up and running, they’ll be able to train a model with 1028 FLOP—a thousand times more than GPT-4. Other companies pour money into their own giant datacenters, hoping to keep pace.
Although models are improving on a wide range of skills, one stands out: OpenBrain focuses on AIs that can speed up AI research. They want to win the twin arms races against China (whose leading company we’ll call “DeepCent”)16 and their US competitors. The more of their research and development (R&D) cycle they can automate, the faster they can go. So when OpenBrain finishes training Agent-1, a new model under internal development, it’s good at many things but great at helping with AI research. By this point “finishes training” is a bit of a misnomer; models are frequently updated to newer versions trained on additional data or partially re-trained to patch some weaknesses.

In our timelines forecast, we predict when OpenBrain will internally develop a superhuman coder (SC): an AI system that can do any coding tasks that the best AGI company engineer does, while being much faster and cheaper.
According to a recent METR’s report, the length of coding tasks AIs can handle, their “time horizon”, doubled every 7 months from 2019 - 2024 and every 4 months from 2024-onward. If the trend continues to speed up, by March 2027 AIs could succeed with 80% reliability on software tasks that would take a skilled human years to complete.
Such is the capability progression in our scenario:
Forecast: Going from 1 week to 1 year mght be ~2x easier than going from 1 hour to 1 week. Reasoning: 1 week tasks can be much more complex than 1 hour tasks, but we project there aren’t as many extra skills needd to go from 1 week to 1 year.

In AI 2027, these capabilities are sufficient for the AI to be an SC, though we have high uncertainty regarding what time horizon might be needed.
In our timelines forecast we combine this time horizon trend with estimations of the gaps between METR’s tasks and the real world to get a distribution for when superhuman coders will arrive. All forecasters place 2027 as one of the most likely years in which an SC might be developed.

Oh wow, das ist super-interessant!

Eine Gruppe Forschender, in den USA, hat skizziert, wie die technologische Entwicklung in Sachen "Künstliche Intelligenz" in den nächsten Jahren, sehr realistisch aussehen kann!

Was wir heute als "Künstliche Intelligenz" bezeichnen, steckt gerade mal in den Kinderschuhen!

Das, was heutige "Large Language Models" können - worüber sich viele hier abfällig lustig machen, weil es ja lediglich stochastisches Aneinanderreihen von Worten sei -, ist vergleichweise "Kindergarten"!

Und die Entwicklung beschreibt gerade mal die nächsten ~ zweieinhalb Jahre!

Nein, das ist kein plumper "diffuser Blick in die Glaskugel", sondern in diesem Longread wird ein tatsächlich sehr realistisches Szenario dargelegt, wie sich das "Technologische Wettrennen zwischen AI-Unternehmen in den USA und China entwickeln kann.

Bezeichnenderweise kommt darin eine europäische KI-Entwicklung nicht vor - als wäre sie nicht existent, nix, nada, wir sind hier wohl lediglich Beobachtende, die am Ende aber die möglichen Auswirkungen voll abbekommen werden - wie jeder Mensch auf diesem Planeten!

Braucht so 1 1/2 Stunden Zeit, sich den Bericht durchzulesen.
Lohnt sich, wenn sich eins dafür interesiert und sich damit beschäftigen möchte - sehr empfehlenswerte Lektüre!

ai-2027.com/

Im nächsten Toot greife ich ein paar Punkte aus diesem sehr umfangreichen Report heraus, weil ich sie für wichtig halte.

#AI2027 #KünstlicheIntelligenz #KI #ArtificialIntelligence #SuperhumanCoder #SuperhumanAIresearcher #SuperintelligentAIresearcher #ArtificialSuperintelligence

Prognóza AI 2027.
Mám z ní rozpačité pocity. Je to až moc pesimistické, počítá to automaticky se scénáři typu Čína ukradne data/modely. A moc přímočaré, svět bývá složitější; zvlášť geopolitika s tím může dost zamávat. O limitech v podobě výpočetního výkonu nebo omezeném množství elektřiny se tam také nemluví.
Je fajn mít scénář, na který bychom si měli dávat pozor, ale nevěřím, že se něco takového stane. Převzetí moci AI už za 5 let mi nepřijde reálné.
https://asociace.ai/prognoza-ai-2027-hlavni-zjisteni-a-varovani-pro-vyvoj-umele-inteligence/
#AI2027

Just finished "AI 2027" — a fascinating take on how AI could reshape our world in just a few years. Definitely worth a read (or listen).
🌐 https://ai-2027.com/

#AI #FutureOfWork #TechTrends #SpeculativeFiction #AI2027

The #meta anti-trust case and an alternative view to #ai2027 are discussed here #ai

https://open.spotify.com/episode/42R3RLEA2sNPLVIu55NV5b?si=ko0XPTLQRDqL5S7TEZQ93A

Fascinating read on the possible scenario predicted by "AI 2027" for the next three years! The prospect of #AI research becoming fully automated could lead to superintelligent systems with goals misaligned with human values. It's a thought-provoking glimpse into a potential future. #AI2027 #AGI

AI 2027

Footnote: The #AI2027 misunderstanding plagues most self-trained (non-academic) #AI researchers as well - successful research is much more "art" than "craft" - you can't brute force it by trying lots of hypotheses - there are just too many. Deep understanding and good intuition are how you find a golden needle in the haystack of researchable hypotheses. #aihype

Have you heard about the AI 2027 forecast? I don't believe it. IMHO the least plausible part is the leap from AI coding to AI research - the story totally underestimates the unimaginably vast spaces of potentially plausible hypotheses that good researchers must use their knowledge and understanding to prune down to hypotheses to actually test. Coding agents are not going to cut it (none available anytime soon.). #aihype #GenAI #LLM #AI2027

相信最先出現的AGI會是AI程序員，緊接的就是AI研究員，後面陸續會有更多的AGI出現代替人類工作。這裏的AGI是指全面超越人類能獨立完成工作的AI。AI程序員目前是差不多了，Aider+Claude3.7的確是超越了八成的人類，至少比我強得多

https://ai-2027.com/

#AI2027

It's amazing to me that both the secular and evangelical cults within MAGA are apocalyptic in nature, with their own versions of "the rapture". For TESCREALists, it's the singularity. For Evangelicals, it's disappearing into the ether to "meet with Jesus"

#rapture #tescreal #evangelicals #maga #musk #eacc #singularity #skynet #project2025 #ai2027 #agi #uspol