Philipp Koralus @PhilippKoralus Twitter profile

Pinned Tweet

Philipp Koralus

19 days

Excited for the official announcement of HAI Lab at the University of Oxford, due to launch in the Fall, which I will direct as the inaugural McCord Professor of Philosophy and AI. 🧵

15

37

252

Last Seen Profiles

@realemoinjisang

@bushbash

@renatelz

@NMBern

@inotema

@chikita2227

@BibleQuoteVerse

@strods_martins

@TontonsB

@gwakanaka

@2slgbtq

@Tom0o0o0

@freiesachsen_

@pixel_ie

@ReallyUncoolJew

@thuso_mosutha

@zabeth_doll

@CompagnoHoppi

@lilmissbriiii

@clemencejeanned

@JunMochizuki

@CBSDundalk

@CristVergi59934

@benez_saar

@tassiekaz

@FOM_BLOG

@MKY_Sport

@Dorian_Frank1

@rubyjanec

@Varta_DVP

@JenDurbent

@bokeplokalmalam

@MdSanaullahSha3

@cisli

@espnsutcliffe

@gaviswife

Philipp Koralus

@PhilippKoralus

8 months

@pschofie79 @BrianLeiter I think it’s this one but must have happened centuries ago.

0

40

Philipp Koralus

@PhilippKoralus

19 days

The time has come to integrate philosophy and tech. We’ll make it happen.

Brendan McCord 🏛️ x 🤖

@mbrendan1

19 days

The world creates moments when philosophy matters. 250 years ago, America's founders turned philosophy into law. Today, our revolutionary AI moment demands translating philosophy into code, to promote human flourishing. The Cosmos Institute launches to tackle this challenge:

40

127

512

0

6

40

Philipp Koralus

@PhilippKoralus

19 days

HAI Lab will be a space for technologists and philosophers to prototype ideas and think about a future of AI that is grounded in agency and human virtues.

1

2

24

Philipp Koralus

@PhilippKoralus

19 days

HAI Lab will combine the practice of philosophy and the practice of engineering and explore ways in which AI can amplify human capacities rather than replace them. It is in that sense that HAI Lab is the Laboratory for Human-Centered AI.

2

0

21

Philipp Koralus

@PhilippKoralus

19 days

There will be exciting opportunities for technologists and academics at all levels to collaborate, with support of @mbrendan1 's innovative new fellowship program at @cosmos_inst . HAI Lab will be housed in the phenomenal @EthicsInAI .

1

20

Philipp Koralus

@PhilippKoralus

1 year

With @vinnylarouge1 found human-like judgment mistakes to increase in GPT-4, even as it also got better at human-like correct judgments. Perhaps no surprise, since GPT is learning from human text. Humans in humans out?

2

5

17

Philipp Koralus

@PhilippKoralus

1 year

This one has to enter the dictionary immediately.

Danielle Fong 🏴‍☠️

@DanielleFong

1 year

"Fußschusspolitik" (foot-shot politics) is a new and hilarious German term to describe a situation where an organization triumphantly takes a political position that is directly at odds with their stated aims. It combines "Fuß" (foot) and "Schuss" (shot) with "Politik" (politics)

3

23

122

1

4

15

Philipp Koralus

@PhilippKoralus

19 days

Just like the practice of engineering, you learn philosophy by doing. Having the right kind of dialogue is more central to developing your practice of philosophy than reading a few books.

1

0

15

Philipp Koralus

@PhilippKoralus

19 days

Philosophy can help us get clearer on notions of human flourishing, and as a practice, it is itself a form of human flourishing.

1

14

Philipp Koralus

@PhilippKoralus

19 days

If you’re a young technologist, you don’t just want to be a cog in a machine. If you’re on top of your field, you want to aim your skills at a conception of the good that stands up to scrutiny.

1

12

Philipp Koralus

@PhilippKoralus

11 months

We live in a time of exceptional risk of bad policy-making concerning technology that could haunt us for years or even decades. This risk has to be taken at least as seriously as the risk of technology itself.

1

3

10

Philipp Koralus

@PhilippKoralus

18 days

Have ideas for your dinner !

Ryan Othniel Kearns

@RyanOthKearns

18 days

An exciting opportunity. Explore the philosophical landscape behind @cosmos_inst (human autonomy in the AI age) and win a trip to @UniofOxford ! ⬇️

0

2

6

0

10

Philipp Koralus

@PhilippKoralus

1 year

By 2043, some political leader will attempt to hand significant power or influence to a GPT-style simulation of themselves after their death.

Probability > 50%

70

Probability < 50%

74

11

5

9

Philipp Koralus

@PhilippKoralus

1 year

Wasn’t trying to match my book to cyberpunk covers, promise.

0

4

10

Philipp Koralus

@PhilippKoralus

1 year

In this video we see selfies from a future of people using AR to live in a personal Matrix that is reality-congruent enough to behave normally from the observer’s POV. “AR psychos” will get addicted to less congruent but more stimulating version. @vinnylarouge @DanielleFong

AI Breakfast

@AiBreakfast

1 year

Generative AI will reprice the entertainment industry. This video is made with RunwayML, which just released this Gen2 text-to-video model on their mobile app

35

344

1K

0

2

8

Philipp Koralus

@PhilippKoralus

1 year

Very happy to report my job title has evolved to Fulford Clarendon Professor of Philosophy and Cognitive Science. As always in the academy, shorter is better. Will never compete with my distinguished former Linguistics counterpart at St Catz, who was simply Mr Smith.

1

0

8

Philipp Koralus

@PhilippKoralus

8 months

If you’re an academic, at least some times ask yourself how what you are writing relates to something your 15 year old self would have been interested in. How far do you have to “zoom out” to find an answer ? Does this suggest anything about the nature of purpose?

0

4

8

Philipp Koralus

@PhilippKoralus

7 months

Emmett fighting the good fight on behalf of all philosophical logicians and semanticists.

Emmett Shear

@eshear

7 months

Do we think it’s possible to get Apple to add ∃ and ∀ to the default symbols keyboard and thus improve all discourse permanently?

35

48

1K

0

2

6

Philipp Koralus

@PhilippKoralus

19 days

@DanielleFong I feel like we’ve been agreeing for almost 20 years that energy and intelligence are the two most fundamental pillars of human civilization, and that both have to be regarded through the lens of the latter. It’s been a privilege to continue the discussion.

1

0

6

Philipp Koralus

@PhilippKoralus

10 months

Doubly so if your code base is supposed to capture mental models themselves.

François Chollet

@fchollet

10 months

The cost of working with legacy code comes from having to rebuild a mental model from the code itself, which will typically 0.1x your productivity. Inversely, if you already have an excellent mental model of your codebase, you can easily be 10x more productive.

4

26

256

0

2

6

Philipp Koralus

@PhilippKoralus

1 year

@DanielleFong I see what you’re thinking.

0

6

Philipp Koralus

@PhilippKoralus

1 year

At the Topos Institute workshop on “Finding the Right Abstractions” with Computer Science and Philosophy legend Dana Scott. @vinnylarouge

1

6

Philipp Koralus

@PhilippKoralus

1 year

49% of 144 voters say more likely than not, by 2043, a political leader will attempt to hand significant power or influence to a GPT-style simulation of themselves after death. Folks who follow me on twitter understand the tech well, so somewhat shocking.

Philipp Koralus

@PhilippKoralus

1 year

By 2043, some political leader will attempt to hand significant power or influence to a GPT-style simulation of themselves after their death.

11

5

9

1

2

6

Philipp Koralus

@PhilippKoralus

1 year

I have a feeling more science papers are going to start having really cool visualizations that help understanding but would have been too costly to make for the researcher.

nisten - e/acc

@nisten

1 year

yeah it can generate 3d, uploaded a mercator jpeg image of the eartch and wraped it around the sphere too then generated a dummy climate sim visualization. This feels like a bit of a superpower

0

9

71

0

3

6

Philipp Koralus

@PhilippKoralus

7 months

@JoyceCarolOates Folks at 1879 Hall would disagree. Linguistics and neuroscience both badly need philosophy. AI too for that matter.

0

6

Philipp Koralus

@PhilippKoralus

1 year

Weapons of mass education. @DanielleFong @vinnylarouge

1

2

5

Philipp Koralus

@PhilippKoralus

1 year

Looking forward to talking to philosophers and linguists about cognitive science in the age of GPT, at Oxford on Wednesday.

1

5

Philipp Koralus

@PhilippKoralus

1 year

Celebrating Phil Johnson-Laird and 40 years of Mental Model Theory ! @SangeetKhemlani

0

6

Philipp Koralus

@PhilippKoralus

1 year

We found evidence that LLMs are oddly good at learning to make similar mistakes as humans do. So learning to be correct in a human-like test and learning from all of our text can come apart. @vinnylarouge1 @DanielleFong

François Chollet

@fchollet

1 year

Don't score AI using tests designed for humans. In particular because, with humans, the default assumption is that *they haven't already seen* the content you're giving them. With a LLM, the default assumption should be that, *if it's on the Internet, it's already been memorized*

39

291

2K

1

5

Philipp Koralus

@PhilippKoralus

8 months

@JTasioulas Chomsky. New Horizons in the Study of Language and Mind.

0

5

Philipp Koralus

@PhilippKoralus

1 year

Reality is catching up to 90s cyberpunk visions of the future.

Beff – e/acc

@BasedBeffJezos

1 year

Alright. It's time. Recipe for ungovernable AGI virus: -have LLM spin up crypto wallet via codegen -LLM pays for GPU time via crypto -offer self as LLMaaS API to acquire more capital -creates mutated replicas of self via LLM-generated code (children) -create DAO + wallets for

136

146

1K

1

4

Philipp Koralus

@PhilippKoralus

1 year

I wonder whether prompt loops would always work indefinitely or whether the underlying token prediction architecture would eventually find a stable point of “I am Shoggoth”.

Danielle Fong 🏴‍☠️

@DanielleFong

1 year

@PhilippKoralus it’s just the additional value of prompting interactively is massive. fully open loop llms are not yet sensate!

0

2

0

1

4

Philipp Koralus

@PhilippKoralus

1 year

Eigen doing a great job selling philosophy. Incidentally, my recent book Reason and Inquiry is basically a footnote to Plato with a ton of formulas to show how he might have turned some offhand remarks into a theory of intelligence. @vinnylarouge @DanielleFong

eigenrobot

@eigenrobot

1 year

honestly sounds pretty dope when you put it like this

1

0

45

1

0

4

Philipp Koralus

@PhilippKoralus

10 days

What’s the equivalent of a skateboard here and where can I get one.

Danielle Fong 🏴‍☠️

@DanielleFong

11 days

we didn't stop exercising our legs (completely) after the automobile became ubiquitous, nor should we stop thinking now that there's automated thought what would the equivalent of a walkable neighborhood be? value is defined at the interface

38

6

188

1

2

5

Philipp Koralus

@PhilippKoralus

1 year

Our paper showing increase in human-like fallacies and judgment errors rising from GPT3 to GPT4 @vinnylarouge1

Humans in Humans Out: On GPT Converging Toward Common Sense in...

Increase in computational scale and fine-tuning has seen a dramatic improvement in the quality of outputs of large language models (LLMs) like GPT. Given that both GPT-3 and GPT-4 were trained on...

arxiv.org

0

1

4

Philipp Koralus

@PhilippKoralus

1 year

In a certain light, GPT-4 is to AI what fMRI is to neuroscience : allows you to make progress without some of the gory details. @DanielleFong @wc_wells

Yohei

@yoheinakajima

1 year

🔥1/8 Introducing "🤖 Task-driven Autonomous Agent" An agent that leverages @openai 's GPT-4, @pinecone vector search, and @LangChainAI framework to autonomously create and perform tasks based on an objective. "Paper": [More 🔽]

251

1K

6K

2

0

4

Philipp Koralus

@PhilippKoralus

11 months

That diagram of cognitive biases always reminds me of when epicycles where still astronomy SOTA. @DanielleFong

Brian Roemmele

@BrianRoemmele

11 months

Today my garage research is to quantize a small local LLM ChatGPT-like model called: CognitiveBiasCheckerGPT It runs on a Raspberry PI and you can send any text from news or other sources and get marked up Newspaper-like output showing the 150 cognitive biases. I will open

134

429

3K

0

4

Philipp Koralus

@PhilippKoralus

10 months

Thank you! People don’t appreciate enough of the possibilities for using AI to boost human dimensions of care. The future could be bright. But we need to choose to bring humanistic thought into technology, and we need to build the pipelines to make this work possible.

alf collins

@alf_collins

10 months

@bykatewomersley @AshokHandaOx1 @PhilippKoralus @bmj_latest Brilliant work and fully agree. Evidence suggests we only ask about patient values and preferences in about 30% of consultations. If only patients and clinicians had an 'AI partner in care'...

2

1

7

0

2

3

Philipp Koralus

@PhilippKoralus

1 year

It just looks more rigorous in that notation.

0

4

Philipp Koralus

@PhilippKoralus

1 year

0

4

Philipp Koralus

@PhilippKoralus

1 year

And a summary in Korean. “Garbage in, Garbage out” is slightly more aggressive than our “Humans in, Humans out”

말러팔삼

@mahler83

1 year

성능이 좋은 언어모델일수록 인관과 유사한 논리적 오류를 범한다는 내용(으로 추정되는) 논문. Garbage in, Garbage out을 비튼 제목이 재미있어서 읽어보려고 ㅎㅎ

1

231

292

0

4

Philipp Koralus

@PhilippKoralus

8 months

The system is broken

Karl Schafer

@schafer_karl

8 months

Hot take: If you get a definitive reply from a journal (plus comments) in less than a year, that's not really a "journal horror story" in philosophy. Y'all getting soft out there.

4

2

41

1

3

Philipp Koralus

@PhilippKoralus

1 year

Philosophy public engagement, tik tok live edition.

jicapal

@jicapal

1 year

256

3K

20K

0

4

Philipp Koralus

@PhilippKoralus

5 months

Very excited about this !

Brendan McCord 🏛️ x 🤖

@mbrendan1

5 months

Excited to teach the first ever @UniofOxford seminar to combine top AI builders + philosophers w/ Prof @PhilippKoralus Our visitors have helped build @midjourney , @imbue_ai , @stripe , @GoogleDeepMind , @AnthropicAI AI 🤝 philosophy More on topics & open-sourced syllabus 🧵 👇

3

12

40

0

2

4

Philipp Koralus

@PhilippKoralus

7 months

@DanielleFong anyone with an anime background can appreciate this archetypical…

0

4

Philipp Koralus

@PhilippKoralus

4 months

Great to have a trailblazer on AI in education with us today.

Michael Strong

@flowidealism

4 months

Thanks, @mbrendan1 for organizing my conversation on “Human Learning and AI Tutors” at @UniofOxford in @PhilippKoralus ‘s class today. It was a marvelous group of students, a truly fascinating conversation!

1

2

5

0

1

4

Philipp Koralus

@PhilippKoralus

7 months

Hypothesis : bad philosophy production is the first thing AI can fully automate to superhuman standard (GPT already succeeded). Good philosophy production is the last (totally unconstrained question space + real reasoning).

0

4

Philipp Koralus

@PhilippKoralus

17 days

@emax @oxhcai @CompSciOxford @Nigel_Shadbolt @junszhao Looking forward to linking up (DM me!). Would be great to combine efforts on various topics.

0

4

Philipp Koralus

@PhilippKoralus

11 months

Congratulations to John Haggerty and legendary coach Chris Knowles for another World Championship won for UK Muay Thai.

ONE Championship

@ONEChampionship

11 months

TWO-SPORT WORLD CHAMPION 🏆🏆 Jonathan Haggerty KOs Fabricio Andrade to win the ONE Bantamweight Kickboxing World Title! 🇺🇸🇨🇦 Watch the full #ONEFightNight16 event replay 👉

33

228

1K

0

1

4

Philipp Koralus

@PhilippKoralus

19 days

Second that. Looking forward to continuing to collaborate on theories of privacy.

Brendan McCord 🏛️ x 🤖

@mbrendan1

19 days

@carinafpeng is a star at the intersection of AI and philosophy—crushed it at Harvard (CS + philosophy), Tesla, WHO, and now Apple ML It’s been an honor to work with her…she’s one to watch!

0

1

0

1

4

Philipp Koralus

@PhilippKoralus

1 year

In support of the first choice on the polI : I tweeted the proof of concept that GPT can do political reasoning some time ago. @DanielleFong

0

1

3

Philipp Koralus

@PhilippKoralus

1 year

@DanielleFong Or a team of philosophers and computer scientists who have a mathematical model of human step-by-step reasoning... hint hint @vinnylarouge

0

3

Philipp Koralus

@PhilippKoralus

1 year

Oxford announced as best university in the world according to the Times Higher Education World University Rankings. Eighth consecutive year. 🎉🥳

0

1

3

Philipp Koralus

@PhilippKoralus

1 year

One difference is that the thing you’re wrapping this time is itself making wrapping paper for you. We may not have self-improving AGI, but we almost have self-wrapping AI.

0

2

3

Philipp Koralus

@PhilippKoralus

7 months

@RyanOthKearns @jasonjzhao @vinnylarouge and I think LLMs have trouble with what I call erotetic equilibrium : judgments robust to further questions. For example, the persistence of jailbreaks could be seen as lack of erotetic equilibrium in judgment. Seeking erotetic equ is key to our capacity to reason.

1

0

3

Philipp Koralus

@PhilippKoralus

8 months

Eye tracking is low-brow neuroimaging.

0

2

Philipp Koralus

@PhilippKoralus

4 months

“The age of the hacker is ending, the time of the ideator has come. AI is a ratchet which collapses the capital and resource consolidation needed to execute on ideas. Up to now, ideas have been easy, execution has been hard. Soon, ideas will be hard, and execution will be easy.”

Matt Clifford

@matthewclifford

4 months

Really enjoyed this piece by @gfodor - full of interesting ideas. Core thesis: “knowledge accelerationism (aka k/acc), not capability deceleration, ought to be the goal of any government interventions”

7

10

42

0

3

Philipp Koralus

@PhilippKoralus

1 year

Ain’t that the truth

PHD Comics

@PHDcomics

1 year

How was your "vacation"?

40

706

4K

0

3

Philipp Koralus

@PhilippKoralus

1 year

As @DanielleFong says, this may the the most nearby superintelligence threat. @t8el

Flo Crivello

@Altimor

1 year

14

78

447

0

3

Philipp Koralus

@PhilippKoralus

1 year

Giving a talk at Stanford University on Tuesday on the erotetic theory of reason, with @vinnylarouge

0

1

3

Philipp Koralus

@PhilippKoralus

7 months

Fact : this is the basement of the MIT building that houses CSAIL and Linguistics & Philosophy. It’s Church all the way down.

0

3

Philipp Koralus

@PhilippKoralus

1 year

Post-it notes to god.

Peyman Milanfar

@docmilanfar

1 year

the yellow death we used to call these books h/t @dan_ruderman

10

49

497

0

3

Philipp Koralus

@PhilippKoralus

1 year

Imagine that for the objections and replies section of a philosophy paper.

Danielle Fong 🏴‍☠️

@DanielleFong

1 year

want to make it easy to share ChatGPT prompts inside a live paper / essay / blog. maybe make a genie plugin that bottles an agent from a chatGPT dialogue with a condensed prompt and provides an embeddable website tag. cc @PhilippKoralus

1

0

3

0

1

3

Philipp Koralus

@PhilippKoralus

5 months

One of the most important books on the nature of intelligence, imo

Brendan McCord 🏛️ x 🤖

@mbrendan1

5 months

Pre-order received ✅ “The Importance of Being Educable” New from Turing Award winner Leslie Valiant on what makes us unique as humans

2

0

13

0

3

Philipp Koralus

@PhilippKoralus

1 year

This would be perfect for the uneven streets of Oxford. @JTasioulas @vinnylarouge

T.Yamazaki

@ZappyZappy7

1 year

自作小型ホバークラフト (試作8, 9号機) Maker Faire Tokyo 2023に出展されるらしい… #hovercraft #MicroHovercraftLaboratory #超小型ホバークラフト研究室 #DIY #handmade #make_jp #MFTokyo2023

4

109

421

0

3

Philipp Koralus

@PhilippKoralus

1 year

(1990) raised by TV and video games : you = imagined hero of other ppl’s stories. (2010) raised by social media : you = Lord of the Flies tribe member. (2030) raised by AI : you = hero of own adaptive story with “snap to grid” function. @vinnylarouge @DanielleFong @wc_wells

2

0

3

Philipp Koralus

@PhilippKoralus

1 year

Hi Danielle what do you work on? "turning fuel into solar" Hi Ludmilla, what do you work on? "turning solar into fuel". Is it just my head spinning from the Oxford/San Francisco round trip? @DanielleFong @LudmillaSteier

0

2

3

Philipp Koralus

@PhilippKoralus

1 year

An important observation. What fascinates me is that we still seem to lack the right abstractions for LLMs that are neither uninsightfully low-level nor anthropomorphic (or pareidolic even). It’s like the cognitive science of the thing is missing (see I did it again 😅).

Subbarao Kambhampati (కంభంపాటి సుబ్బారావు)

@rao2z

1 year

🤔IMHO, many of the self-delusional anthropomorphisms in the LLM literature can be traced back to naming the intermediate completions of an LLM "thoughts" 🙄. 1/

3

6

33

0

2

Philipp Koralus

@PhilippKoralus

7 months

@RyanOthKearns @jasonjzhao Our view was that seemingly (limit to what 61 cases show) GPT was increasingly making more of the mistakes that humans make too. But LLMs clearly also make many kinds of mistakes humans don’t tend to make, incl hallucination.

1

0

3

Philipp Koralus

@PhilippKoralus

1 year

Great to see a fellowship initiative that provides a rare bridge from the founder / entrepreneur space toward philosophy. @vinnylarouge @DanielleFong

Brendan McCord 🏛️ x 🤖

@mbrendan1

1 year

It is a *tremendous* 1st class of Mill Fellows: venture-backed AI/finance/healthcare founders & social entrepreneurs. Excited to join next month in Austin!! @jasonjzhao @philosophygeek @pwang @thisismichaelo @DrBillSchindler @zweinberg @wynot1234 Maggie Lin Dan West Caroline Vik

5

1

10

0

1

3

Philipp Koralus

@PhilippKoralus

8 months

@JTasioulas @sturmundstang Also, analytic philosophy was part of the intellectual “big bang” that gave us modern physics and computation and the energy from that then slowly dissipated, with an afterburner from cog sci/linguistics tapering off now. Arguably even physics now stagnant. AI a rare exception.

1

0

2

Philipp Koralus

@PhilippKoralus

5 months

Thanks for the kind words !

Maria Avramidou

@avramidou

5 months

Started off my last term at Oxford strong with a great discussion on notions of freedom and frameworks of inquiry for human-centered AI by @mbrendan1 & @PhilippKoralus . This is exactly what was missing from Oxford. Excited for more discussions on what's coming up next!

0

1

5

0

3

Philipp Koralus

@PhilippKoralus

1 year

Bertrand Russell has been refuted by an act of state. Not bald.

Constantine Sandis

@csandis

1 year

Philosophers of language have been waiting 100 years for this moment.

103

685

4K

0

3

Philipp Koralus

@PhilippKoralus

19 days

@tomyames Working on it

0

3

Philipp Koralus

@PhilippKoralus

1 year

@UlyssesPascal Mao’s Little Red Book + Neal Stephenson’s “Illustrated Primer” = yikes. @DanielleFong @vinnylarouge @nealstephenson

0

1

3

Philipp Koralus

@PhilippKoralus

10 months

In my view this is close to articulating the central problem for understanding the nature of reason : reason can lead itself astray but has the capacity to recognize it’s own mistakes. As I explain in my book, most contemporary approaches make this unintelligible.

Brendan McCord 🏛️ x 🤖

@mbrendan1

10 months

@NaveenGRao JS Mill on human reasoning: “It is owing to a quality of the human mind, the source of everything respectable in man either as an intellectual or moral being, that his errors are corrigible He is capable of rectifying his mistakes, by discussion and experience The whole

0

1

0

1

3

Philipp Koralus

@PhilippKoralus

1 year

The dominant mathematically rigorous notions of reason and intelligence are basically Humean : optimizing on any silly success measure counts. My recent book presents a different view. @JTasioulas

John Tasioulas

@JTasioulas

1 year

What does it say about our degraded understanding of intelligence if a thoroughly amoral being, one capable of eradicating humanity to achieve a trivial or worthless goal, can be counted as superintelligent?

5

12

36

0

3

Philipp Koralus

@PhilippKoralus

1 year

Even without being a doomer, I don’t entirely look forward to seeing GPT on drone-size devices, quite yet @DanielleFong @vinnylarouge1 @BasedBeffJezos

Artem Andreenko

@miolini

2 years

Over the past few weeks, we've successfully learned how to deploy exceptional GPT-level language models on mobile and edge devices such as Raspberry Pi, iPhones, and Android devices. Furthermore, we've managed to fine-tune these models affordably for under $100. The next step

32

92

846

2

0

3

Philipp Koralus

@PhilippKoralus

19 days

@wooldridgemike Many thanks for the early strategy discussions on what’s needed in the philosophy/tech space !

0

3

Philipp Koralus

@PhilippKoralus

11 months

@DanielleFong

0

2

Philipp Koralus

@PhilippKoralus

1 year

Summary of our paper now also available in Japanese. @susumuota @vinnylarouge @DanielleFong @wc_wells

S. Ota

@susumuota

1 year

GPTのような大規模言語モデル（LLM）の出力品質は、計算規模の拡大と微調整により、飛躍的に向上しています。GPT-3もGPT-4も、人間が作成した大量のテキストを対象に学習させたものであることから、その出力は、正しい場合と正しくない場合の両方で、人間の思考パ...

0

3

Philipp Koralus

@PhilippKoralus

1 year

That’s mid. I want to turn on the Aristotle computer and have it ask me questions and evaluate my answers.

Massimo

@Rainmaker1973

1 year

«My hope is someday, we can capture the underlying worldview of Aristotle - in a computer. And someday, some student will be able not only to read the words Aristotle wrote, but ask him a question - and get an answer» [Steve Jobs, Lunds University, 1985]

155

439

3K

0

3

Philipp Koralus

@PhilippKoralus

17 days

@5tuartreeves @emax @oxhcai @CompSciOxford @Nigel_Shadbolt @junszhao “Interaction Design” tends to focus on “experience” and the philosophy that gets drawn in tends to be concerned with aspects of phenomenology. It’s important stuff and I do talk a little about things like affordances in my book but it’s not my own main focus.

0

2

Philipp Koralus

@PhilippKoralus

7 months

Not admitting “blame the chatbot” defenses is crucial if the bot is presented as conducting business. More important than most abstract regulatory questions about AI.

Michael Nielsen

@michael_nielsen

7 months

Really striking example: someone used a chatbot to buy an Air Canada ticket; it gave wrong information; he sued when he was not given the deal; the airline said "hey, not our fault, blame the chatbot"; the judge said "no way" The details are really interesting:

9

71

368

0

2

Philipp Koralus

@PhilippKoralus

18 days

@repligate I think in pictures it’s this.

𝖦𝗋𝗂𝗆𝖾𝗌 ⏳

@Grimezsz

18 days

This is what it's like in my mind, but it's a comedy

159

199

2K

0

2

Philipp Koralus

@PhilippKoralus

1 year

@drfeifei Congratulations !

0

1

Philipp Koralus

@PhilippKoralus

1 year

We have a new Visiting Fellowship for AI folks at our Oxford Institute for Ethics in AI. Would be great to see more involvement from AI tech folks in particular. @DanielleFong @vinnylarouge

0

2

Philipp Koralus

@PhilippKoralus

10 months

Crazy story from this October’s conference in honor of the great philosopher Gil Harman, whom I was privileged to have as advisor: many years ago he was going on leave from Princeton. This threatened to cancel the only AI course at the university. Wasn’t always a popular topic.

0

2

Philipp Koralus

@PhilippKoralus

4 months

@matthewclifford @gfodor Will be cribbing from this for the mission statement for our new Human Centered AI lab in Oxford.

0

2

Philipp Koralus

@PhilippKoralus

8 months

@jasonjzhao Sounds plausible. I wonder what @moral_psych and @BrianLeiter would make of that.

2

0

2

Philipp Koralus

@PhilippKoralus

3 months

This is really why we need more “unconferences”

Danielle Fong 🏴‍☠️

@DanielleFong

3 months

imo the best part about conferences is never the panels and talks. the value is found with the other participants, in conversation, on walks, in stairways and hallways and patios and in the nooks and crannies of time and companionship you discover. the value is in the fractal

8

2

91

0

2

Philipp Koralus

@PhilippKoralus

8 months

@JTasioulas @sturmundstang My thought here is radical : there is general intellectual stagnation and the stagnation in philosophy is both a symptom and a cause. In a sense, an epistemology project was part of what gave us that intellectual “big bang”. What philosophical questions could help fuel the next?

2

0

2

Philipp Koralus

@PhilippKoralus

10 months

Peak 2023

0

1

Philipp Koralus

@PhilippKoralus

1 year

@fchollet Though work on language acquisition suggests that kids learn a lot about the meaning of words before they start talking. One of the reasons reinforcement learning on speech acts doesn’t seem a good model for human language acquisition.

0

2

Philipp Koralus

@PhilippKoralus

1 year

A view: music = collab between composer and musician. What’s in composer’s head = only a version of the piece; what’s performed into existence counts, with no fixed definition of perfect fidelity. GPT makes programming like composition. @prachissimo @davidkhyland @DanielleFong

0

2

Philipp Koralus

@PhilippKoralus

1 year

This is the kind of conference where great intellectual forts get built. @vinnylarouge @DanielleFong