Carissa Véliz on AI
6 quotes from Carissa Véliz on the wall, 2023 to 2026, newest first. Each one is verbatim and links to where it was said. Sources include EL PAÍS English, BBC Science Focus Magazine, CNET.
- 2026
It doesn't understand the world because it doesn't inhabit it
Dr Carissa Véliz, an AI ethicist at the University of Oxford, explaining in BBC Science Focus why AI chatbots produce factual errors.
What it refers to
This explains why AI chatbots in a BBC study produced factual errors and fabricated quotes about news stories.
Other voices
- AI can imitate and simulate, but it cannot educateFr. Thomas Dailey28 August 2026
- Are not your friends, are not conscious beingsMeredith WhittakerBuilder21 June 2026
- It is hard and perhaps impossible to separate what they do from what they areAnil SethResearcher15 July 2026
- If AI has consciousness, then we need to treat it as a moral subject is the wrong way of thinkingBlaise Agüera y ArcasResearcher30 August 2026
- How many selves does AI have?Markus GabrielThinker10 April 2026
BBC Science Focus Magazine7 ways we're trusting AI too much – and already paying for it
When we hear a prediction and take it as fact, what we're really doing is obeying
Carissa Véliz, Oxford philosopher, discussing how AI-driven predictions function as instruments of power, EL PAÍS interview, May 2026.
What it refers to
AI-driven predictions presented as factual descriptions rather than as commands that shape future behavior and reinforce power structures.
Predictions are often commands disguised as a search for knowledge
Carissa Véliz, Oxford philosopher and author of Prophecy, interviewed by EL PAÍS in May 2026 about predictions, statistics and AI.
What it refers to
Predictions treated as neutral facts that actually shape expectations and behavior, a theme explored in her book Prophecy.
Other voices
- The state of the field feels like physics before NewtonWilliam MacAskillThinker20 July 2026
- The longtermists' spreadsheet becomes civilization itselfDavid MillsResearcher19 May 2026
- We are all going to die within a few decades anywayNick BostromThinker17 June 2026
From Carissa Véliz's CNET essay (2026), adapted from her book Prophecy, on how machine learning extends humanity's long history of prediction and power.
Other voices
- Make AI look like magicAnil SethResearcher15 July 2026
- I don't think AGI should be turned into a marketing term for commercial gainDemis HassabisBuilder29 January 2026
- The only way to push back against the giant tsunami of AI hype we're all facing right nowAngèle Christin15 June 2026
- While the company warns of power-seeking AI, it has itself become the power-seeking entityNirit Weiss-Blatt3 March 2026
- As far as putting words in order is concerned, AI already thinks better than many of usYuval Noah HarariThinker28 January 2026
CNETFrom Tea Leaves to AI: Why Today’s High-Tech Predictions Are So Dangerous
- 2025
There is a fear that we need to keep up the hype, or else the bubble might burst
Carissa Véliz, Institute for Ethics in AI, commenting on OpenAI's GPT-5 launch, August 2025.
What it refers to
GPT-5 was marketed by OpenAI as having PhD-level expertise and marking a new era for ChatGPT.
Other voices
- When bubbles happen, smart people get overexcited about a kernel of truthSam AltmanBuilder15 August 2025
- It's amazing how much a PhD level machine can get wrongHarry Shearer31 August 2025
- Practically the only people still pretending scaling LLMs is all you need are griftersGary MarcusResearcher26 September 2025
- I am completely unimpressed by LLMs and not worried about AGIDavid Sun30 October 2025
- Even the best systems available today lack anything that could be called intelligenceEmily BenderResearcher30 June 2025
BBC NewsOpenAI claims GPT-5 model boosts ChatGPT to 'PhD level' - BBC
- 2023
We're definitely not ready for what's coming
Carissa Véliz, philosopher at the University of Oxford, commenting to New Scientist in February 2023 after a Google Bard advert showed the chatbot giving a factually wrong answer.
What it refers to
An advert for Google's AI chatbot Bard was shown giving a factually wrong answer about a James Webb Space Telescope discovery.
Other voices
- Not ready for our post-truth information overlordsGary MarcusResearcher30 December 2022
- Not something that people can use reliably on a daily basisZoubin Ghahramani21 December 2022
- I would advocate not moving fast and breaking thingsDemis HassabisBuilder28 April 2023
- We should be really terrified of this whole thingTimnit GebruResearcher22 January 2023
- Know what it knows, and know what it doesn't knowHuang Mingjie15 December 2022
New ScientistGoogle Bard advert shows new AI search tool making a factual error