Sign in

Dr Abeba Birhane

@abeba.blacksky.app
54K followers 915 following 5K posts

Founder & PI @aial.ie. Assistant Professor of AI, School of Computer Science & Statistics, @tcddublin.bsky.social AI accountability, AI audits & evaluation. Cognitive scientist by training. Ethiopian in Ireland. She/her

PostsRepliesMedia
Dr Abeba Birhane @abeba.blacksky.app · 20/09/2026
i hate it @wileyonlinelibrary.bsky.social, please kill this
2548779
Dr Abeba Birhane @abeba.blacksky.app · 14/09/2026
a scooby doo "unmasking" meme. it reads: 

highly capable, powerful, advanced super-intelligent AI 

army of humans hired by OpenAI to read real ChatGPT users' chats
219680
Dr Abeba Birhane @abeba.blacksky.app · 12/09/2026
እንብላ።
5882
Dr Abeba Birhane @abeba.blacksky.app · 12/09/2026
the trash talker On South Anne Street in Dublin city centre, a satirical statue of Donald Trump in a wheelie bin with a banana on his head titled 'The Trash Talker' has appeared. www.rte.ie/news/post/10...
111036
Dr Abeba Birhane @abeba.blacksky.app · 11/09/2026
just this... currently cooking :)
160
Dr Abeba Birhane @abeba.blacksky.app · 11/09/2026
yup, and Joseph Weizenbaum articulated this playbook clearly back in 1972 "This is usually coupled to the more or less subtle assertion that only computer science, hence only the computer scientist, can guard the world against the admittedly hazardous fallout of applied computer technology."
a screenshot reads: The structure of the typical essay
on "The impact of computers on so-
ciety" is as follows: First there is an
"on the one hand" statement. It tells
all the good things oomputers have
already done for society and often even
attempts to argue that the social order
would already have collapsed were it
not for the "computer revolution."
This is usually followed by an "on the
other hand" caution which tells of cer-
tain problems the introduction of com-
puters brings in its wake. The threat
posed to individual privacy by large
data banks and the danger of large-
scale unemployment induced by indus-
trial automation are usually mentioned.
Finally, the glorious present and pro-
spective achievements of the computer
are applauded, while the *dangers al-
luded to in the second part are shown
to be capable of being alleviated by
sophisticated technological fixes. The
closing paragraph consists of a plea
for generous societal support for more,
and more large-scale, computer re-
search and development. This is
usually coupled to the more or less
subtle assertion that only computer
science, hence only the computer sci-
entist, can guard the world against the
admittedly hazardous fallout of applied
computer technology.
In fact, the computer has had very
considerably less societal impact than
the mass media would lead us to be-
lieve. Certainly, there are enterprises
like space travel that could not have
been undertaken without computers.
Certainly the computer industry, and
with it the computer education indus-
try, has grown to enormous propor-
tions. But much of the industry is
self-serving. It is rather like an island
economy in which the natives make a
living by talking in each other's laundry.
The part that is not self-serving is
15219
Dr Abeba Birhane @abeba.blacksky.app · 09/09/2026
mood
reads: do not talk to me about "AI". you make me long for the sweet embrace of death
18916
Dr Abeba Birhane @abeba.blacksky.app · 07/09/2026
what a pile of trash
These debates, intensifying post-2021 amid broader AI ethics discourse, question whether Birhane's focus on power dynamics and colonial legacies in data systematically undervalues first-principles evaluations of AI's causal benefits, such as efficiency gains from diverse, uncurated training corpora that mirror real-world variability. Critics, including those in machine learning communities, argue this
3120
Dr Abeba Birhane @abeba.blacksky.app · 07/09/2026
racism vs "innovation" apparently a trade-offs that one has to choose from
In the context of Birhane's dataset bias audits, such as her examinations of LAION corpora revealing scaled increases in derogatory content, responses highlight a methodological tilt toward cataloging ideological harms—e.g., counts of slurs or stereotypes—while downplaying trade-offs in model utility and innovation. Peer-reviewed analyses of debiasing interventions, including dataset filtering to mitigate
5220
Dr Abeba Birhane @abeba.blacksky.app · 07/09/2026
it’s true that it is distractive and a waste of time to read everything written about you
Critics have argued that Birhane's advocacy for a relational ethics framework, which emphasizes qualitative assessments drawn from the lived experiences of marginalized communities over rationalist quantitative methods, risks substituting ideological priors for empirical rigor in evaluating AI systems.
5371
Dr Abeba Birhane @abeba.blacksky.app · 04/09/2026
good evening from #Dublin
2651
Dr Abeba Birhane @abeba.blacksky.app · 11/08/2026
"A student’s essay on Macbeth has no value as a product. It’s what you learned in the reading and writing that created it – the swiftness and resilience and suppleness of thinking you built, along with the knowledge gained of a literary marvel, of something unsummarisable – that was the point"
Many students want to resist, but the temptation is great, and some start to feel they’re being penalised for their honesty. I tell the students I meet: imagine you’re an Olympic athlete. You set out to sprint the 3km across London, from Clapham to Balham. You know if you take the bus, it will be faster. But you also know that the point was never being in Clapham or Balham; they’re very similar: they both have a nice Pizza Express. It was the strength you built, the speed and knowledge of your body that you gained. A student’s essay on Macbeth has no value as a product. It’s what you learned in the reading and writing that created it – the swiftness and resilience and suppleness of thinking you built, along with the knowledge gained of a literary marvel, of something quicksilver and unsummarisable – that was the point. So you can’t collaborate with AI in the same way you can’t collaborate with a bus. And you cannot learn to think without a climbing frame of facts and ideas and texts to move around in.
210729
Dr Abeba Birhane @abeba.blacksky.app · 11/08/2026
"AI, we’re told, will make learning more efficient. But is that what we dream of for our kids – an efficient childhood? To write an efficient story, sing an efficient song?"
AI, we’re told, will make learning more efficient. But is that what we dream of for our kids – an efficient childhood? To write an efficient story, sing an efficient song? An adequate teacher is not only teaching about Henry VIII; they’re teaching your child how to read a human character. A good teacher is showing them how to shape an argument, how to command an audience, how to tell jokes, how to pass on knowledge. A great teacher is showing your child what it is to love. They are teaching your child how to be human. A computer will only teach your child how to be a computer.
48926
Dr Abeba Birhane @abeba.blacksky.app · 09/08/2026
we made similar points about AI systems and humans as things that are different in kind 6 years ago and all of the points in our paper still stand Robot Rights?: Let's Talk about Human Welfare Instead dl.acm.org/doi/abs/10.1...
Our starting point is not to deny robots ‘rights’, but to deny
that robots are the kinds of beings that could be granted or denied
rights. We suggest it makes no sense to conceive of robots as slaves,
since ‘slave’ falls in the category of being that robots aren’t. Hu-
man beings are such beings. We believe animals are such beings
(though a discussion of animals lies beyond the scope of this paper).
We take a post-Cartesian, phenomenological view in which being
human means having a lived embodied experience, which itself is
embedded in social practices. Technological artifacts form a crucial
part of this being, yet artifacts themselves are not that same kind
of being. The relation between human and technology is tightly
intertwined, but not symmetrical.
18415
Dr Abeba Birhane @abeba.blacksky.app · 09/08/2026
"One of Anthropic’s defining pieces of research appears in a paper entitled “On the Biology of a Large Language Model”. This is highly misleading: LLMs do not have any biology at all, because they are not alive. But in reading Anthropic’s paper one is invited quietly to forget this;.."
The field of machine learning frequently employs biological language to describe LLMs: “neurons” and “neural networks”, “training” and “pruning”. Generative AI systems, says Anthropic’s Olah, are grown more than they are built. One of Anthropic’s defining pieces of research appears in a paper entitled “On the Biology of a Large Language Model”. This is highly misleading: LLMs do not have any biology at all, because they are not alive. But in reading Anthropic’s paper one is invited quietly to forget this; the word “metaphor” does not appear in the article.
827873
Dr Abeba Birhane @abeba.blacksky.app · 06/08/2026
no matter who is backing it or how much money is behind it, the idea of a “post-agi society” is naive at best and at worst, intentionally ignorant of the backbone of AI: humans AI (AGI, whatever that is) is human through and through 1/
If we develop AGI, what happens next? What good stable states might we hope to reach, and what should we be doing now to prepare? This workshop brings together thinkers from machine learning, economics, political science, sociology, and more to answer an increasingly pressing question: what assumptions about civilization will AGI break, and how can we be ready to fix them?

This iteration of the workshop will focus in particular on the following themes:
Political economy

How can our political systems adapt to a world where humans are no longer necessary? What does the equilibrium look like? Are there equilibria which preserve non-competitive human values?
The role of AI

What roles might AIs take on in society? How will they think about themselves? What does it look like for AIs to be independent economic actors? What forces will shape their dispositions?
Cultural and moral change

Where will AI values and beliefs come from? What happens when humans aren't the only source of ideas? Are there convergent notions of "goodness", and what do they mean for digital beings?
2548
Dr Abeba Birhane @abeba.blacksky.app · 12/07/2026
Alfred Nobel established the Nobel Prize after wake up call: newspapers mistakenly published his obituary when his brother died & what he read as his legacy horrified him speaking ill (factually) of the dead is literally the most effective way to curb living racist & Nazi public figures/politicians
There is a well-known story about the origin of the Nobel Prize, although historians have been unable to verify it, and some dismiss the story as a myth.[26] In 1888, the death of his brother Ludvig supposedly caused several newspapers to publish obituaries of Alfred in error. One French newspaper[which?] condemned him for his invention of military explosives—in many versions of the story, dynamite is quoted, although this was mainly used for civilian applications—and this is said to have brought about his decision to leave a better legacy after his death.[6] The obituary supposedly stated, Le marchand de la mort est mort ("The merchant of death is dead"),[6] and according to the myth went on to say, "Dr. Alfred Nobel, who became rich by finding ways to kill more people faster than ever before, died yesterday."[27] Nobel read the obituary and was appalled at the idea that he would be remembered in this way. His decision to posthumously donate the majority of his wealth to found the Nobel Prize has been credited to him wanting to leave behind a better legacy.[28][6] However, it has been questioned whether or not the obituary in question actually existed.[28] An actual obituary (pictured) stated "A man who can only with the utmost difficulty be considered a benefactor of mankind has died yesterday at Cannes. It is Mr. Nobel, inventor of dynamite. Mr. Nobel was Swedish." It was published on the first page of the Figaro on the 15th of April 1888 and retracted the next day.
32413695
Dr Abeba Birhane @abeba.blacksky.app · 11/07/2026
perpetual war is a business model for guys like Thiel www.ft.com/content/cb53...
Investors are pouring billions into defence technology start-ups as wars in Ukraine and the Gulf turn drones, autonomous vessels and battlefield AI into one of this year’s hottest venture capital bets.
Companies in the sector have raised $12.3bn from VC funds since the start of 2026, almost double the amount raised in the same period last year and more than the $9.95bn achieved in the whole of last year, according to data compiled for the Financial Times by PitchBook.
The surge comes as conflicts expose demand for next-generation weapons systems that are cheaper and faster to produce. However, with some funds willing to pay increasing multiples as they bet heightened government defence spending will endure, it has also raised concerns that parts of the market are overheating.The Gulf conflict has also sharpened interest in maritime defence tech. UK start-up Kraken Technology, whose autonomous minehunting vessels have been picked by the Royal Navy for deployment in the Strait of Hormuz, is seeking to raise about $100mn at a roughly $1bn valuation, according to people familiar with the matter. 
The company is working with advisers at PJT Partners on the process, the people said. Kraken and PJT declined to comment.The boom is heavily concentrated in the US. American start-ups captured $11.4bn of the total — although Anduril Industries alone accounted for almost half of that.
The Silicon Valley-based company, known for its drones and surveillance towers, almost doubled its valuation to $61bn in a $5bn funding round last month, with backers including Thrive Capital and Andreessen Horowitz.
Other US start-ups to have raised capital include autonomous surface vessel specialist Saronic Technologies and aerial drone maker Shield AI.
07629
Dr Abeba Birhane @abeba.blacksky.app · 08/07/2026
a friend currently attending the ai for good conference sent this and I'm not sure how to feel lol
0262
Dr Abeba Birhane @abeba.blacksky.app · 30/06/2026
misinterpretation of laws a recurring tactic of Big AI read our paper: dl.acm.org/doi/10.1145/...
Table 1. Taxonomy of capture mechanisms with brief descriptions. The highlighted concepts denote five broad, high-level
categories, each further comprising a set of detailed mechanism categories and descriptions.
23813
Dr Abeba Birhane @abeba.blacksky.app · 30/06/2026
000
Dr Abeba Birhane @abeba.blacksky.app · 28/06/2026
we put forward a list of recommendations including for scientific publication and conference venues including #FAccT to develop mechanisms requiring declaration of conflict of interest from Big AI #FAccT2026
1142
Dr Abeba Birhane @abeba.blacksky.app · 28/06/2026
we analyse 100 media articles clustered around global significant events (the UK, Korea and French AI Summits) and find both regulatory and narrative capture across the board #FAccT2026
1134
Dr Abeba Birhane @abeba.blacksky.app · 28/06/2026
we develop a taxonomy of 27 mechanisms used to capture rule making and public disclosure #FAccT2026
1145
Dr Abeba Birhane @abeba.blacksky.app · 28/06/2026
@storitu.org presents the paper now starting with our definition of "Big AI" #FAccT2026
285
Dr Abeba Birhane @abeba.blacksky.app · 27/06/2026
Suchman's excellent keynote zooms in into modern warfare & the role of AI in the Gaza genocide #FAccT2026 "The most consequential decision are made by systems that cannot explain themselves, supplied by companies that answer to no-one, in conflicts that generate no accountability and no reckoning"
"The most consequential decision are made by systems that cannot explain themselves, supplied by companies that answer to no-one, in conflicts that generate no accountability and no reckoning"
18127
Dr Abeba Birhane @abeba.blacksky.app · 27/06/2026
"The us military budget exceeds the next 9 countries combined" Lucy Suchman #FAccT2026
2323
Dr Abeba Birhane @abeba.blacksky.app · 27/06/2026
In her keynote, Lucy Suchman reminds us that raw data is an oxymoron and a bad idea. #FAccT2026
49519
Dr Abeba Birhane @abeba.blacksky.app · 26/06/2026
"Ideas such as UBI, the American sovereign wealth fund & other social welfare distribution systems associated with AGI presume that knowledge of human future falls on the shoulders of few technocarats rather than ideas that need to be put forward for democratic deliberation" Shazeda Ahmed #FAccT2026
05014
Dr Abeba Birhane @abeba.blacksky.app · 26/06/2026
"AGI, the automate everything machine, is a business pitch that is extremely lucrative to companies like anthropic" @bcmerchant.bsky.social in a very humorous craft session at #FAccT2026
a presentation slide. It reads "agi and the new luddism" image of Brian Merchant on the corner
1585
Dr Abeba Birhane @abeba.blacksky.app · 26/06/2026
Dick Blankvoort presents work that looks into model providers obligation to provide a public summary of training data for GPAI models under the EU AI Act: tldr small open source model providers are following rules while there's no information to be found on large providers such as openai #FAccT2026
1134
Dr Abeba Birhane @abeba.blacksky.app · 26/06/2026
Mcllwain now tells the story of Black women early computer visionaries. he describes Julia Beale Williams imaginary 1980s piece, "Computer Magic" and warns "but you won't find her work on the Internet" #FAccT2026
13310
Dr Abeba Birhane @abeba.blacksky.app · 26/06/2026
Charlton McIlwain makes such a poignant point at #FAccT2026: what has the decades of work that has measured, documented, & evaluated the failure and bias of AI systems done? we have known the negative impact for decades yet very little has changed
39233
Dr Abeba Birhane @abeba.blacksky.app · 18/06/2026
good morning #liffey
The river liffey, Dublin. Dramatic morning clouds, calm waters showing reflection of buildings from both sides of the river
0344
Dr Abeba Birhane @abeba.blacksky.app · 17/06/2026
dead funny but also painfully accurate
28620
Dr Abeba Birhane @abeba.blacksky.app · 13/06/2026
also, c'mon...
0140
Dr Abeba Birhane @abeba.blacksky.app · 12/06/2026
Most kids overestimate what AI can do, even those who think they know a lot about how it works. 7/
15011
Dr Abeba Birhane @abeba.blacksky.app · 12/06/2026
One in six kids who have used AI chatbots say an AI has shown or said something they felt “wasn’t OK or wasn’t meant for someone their age”—and most never told a trusted adult about it. 6/
313154
Dr Abeba Birhane @abeba.blacksky.app · 12/06/2026
Among kids who have used AI to discuss their feelings or personal problems (37% of kids who use AI), one in four (25%) say they sometimes feel AI understands them better than most people do. 5/
1379
Dr Abeba Birhane @abeba.blacksky.app · 12/06/2026
Kids who report feeling lonelier tend to use AI more frequently & use AI more often for social & emotional support half of kids who use/interact with AI daily (54%) or weekly (47%), & 44% of monthly AI users, say they feel lonely at least some of the time, compared to 33% who use AI < once a month
25519
Dr Abeba Birhane @abeba.blacksky.app · 12/06/2026
Among kids who use AI for homework help (85%), 16% say they have difficulty starting or completing their homework when they cannot use AI, and 28% of kids who use AI for homework on a daily basis say it is hard when they cannot use it 3/
37821
Dr Abeba Birhane @abeba.blacksky.app · 12/06/2026
Among kids who use AI, 89% say they use it for fun or entertainment, 85% use it for help with schoolwork or homework. 3 in 4 kids (75%) also say they use AI to create things. 48% say they use AI to help with schoolwork at least once a week, including 21% who use it for schoolwork every day. 2/
2458
Dr Abeba Birhane @abeba.blacksky.app · 12/06/2026
one of the most comprehensive surveys on children’s use of AI--conducted March 18-26, 2026, among 9-17yr olds in the US (N = 1,204)--has been released www.commonsensemedia.org/sites/defaul... highlighting key findings 🧵 "Nearly 9 in 10 kids age 9 to 17 (86%) use/interact with AI"
Nearly 9 in 10 kids age 9 to 17 (86%) use or interact
with AI, including about one in four (24%) who do so
on a daily basis.
Teens use AI more often, but use is widespread even among younger
adolescents, with 81% of 9- to 12-year-olds, 89% of 13- to 15-year-olds, and
92% of 16- to 17-year-olds saying they use or interact with AI. Nearly 3 in 10
teens (29% of 13- to 17-year-olds) and 17% of 9- to 12-year-olds use AI daily.
AI-generated summaries in search engine results are the most widely used
form of AI among 9- to 17-year-olds, with 75% having used or interacted with
them. Just over two-thirds (67%) have used AI chatbots, including 58% of 9- to
12-year-olds, 71% of 13- to 15-year-olds, and 77% of 16- to 17-year-olds.
However, when kids are asked which AI they use most often, ChatGPT is by far
the most frequently mentioned (by 40% of all 9- to 17-year-olds), followed by
Google’s AI offerings, including Gemini, AI summaries, or other Google AI
tools (24%).
917486
Dr Abeba Birhane @abeba.blacksky.app · 10/06/2026
remember this ?!
Facebook's artificial intelligence robots shut down after they start talking to each other in their own language

'you i i i everything else'
Andrew Griffin
Thursday 10 September 2020 15:38 BST
1282
Dr Abeba Birhane @abeba.blacksky.app · 10/06/2026
The Irish State Examinations Commission used an AI generated image in one of the English papers for the 2026 leaving cert (Irish college entrance exams) www.examinations.ie/archive/exam...
7309
Dr Abeba Birhane @abeba.blacksky.app · 05/06/2026
look at my peonies
light pink peonies pictured from the top capturing the full blossom
1327916
Dr Abeba Birhane @abeba.blacksky.app · 05/06/2026
anthropic: we have no way of controlling our fully autonomous, "recursively self-improving" model hidden gig workers, RLHF annotators, and seemingly invisible coders & academics that are *actually* improving the model: hold my beer africa.businessinsider.com/news/inside-...
 Inside the unseen operation to turbocharge Claude Code

Shubhangi Goel
01 June 2026 05:02 AM

An Anthropic project is using feedback from about 1,000 human software engineers to improve the performance of Claude Code,the AI coding tool whose recent advancements have disrupted the vibe-coding industry.
522071
Dr Abeba Birhane @abeba.blacksky.app · 05/06/2026
worth reemphasing who palantir is
210778
Dr Abeba Birhane @abeba.blacksky.app · 04/06/2026
what stage of voluntary rights-rendering-to-surveillance is this
Opt-In Surveillance Is Approaching

AIs with access to all our data will soon be able to vouch for us to others. As people come to trust AI judgments of character, not sharing one will look suspicious.
Steven Veld
on
Jun 3, 2026
34413
Dr Abeba Birhane @abeba.blacksky.app · 04/06/2026
what a great privilege to be part of a panel “Shaping the Future: Young People and Artificial Intelligence” this morning at launch event of this hefty report on AI and youth by the National Youth Council of Ireland. the kids are alright!
4569