Scroll to dive in
Good AnalysisContactGood Analysis · AI
Sixty years of the moments AI stopped being a demo and started changing what people believe, lose and fear.
1966
ELIZA was a few hundred lines of pattern matching at MIT. People still poured their hearts out to it. Its creator, Joseph Weizenbaum, wrote that his own secretary asked him to leave the room so she could talk to it in private.
Weizenbaum, Computer Power and Human Reason, 1976; CACM, January 1966
16h
Microsoft launched Tay, a chatbot that learned from the people talking to it on Twitter. A coordinated group taught it to post racist and abusive replies. Microsoft pulled it offline within sixteen hours and apologised.
4–1
DeepMind’s AlphaGo beat Lee Sedol, one of the best Go players alive, four games to one. Three years later Lee retired from professional play, saying that even as number one he would not be at the top, because AI could not be beaten.
2018
An Uber self-driving test car struck and killed Elaine Herzberg as she crossed a road in Tempe, Arizona. Investigators found the system never settled on what she was. The backup driver was charged; the company was not.
2022
Blake Lemoine, an engineer at Google, went public with his belief that the company’s LaMDA chatbot was sentient. Google, which had already placed him on leave, said the evidence did not support it and fired him the following month.
100M
OpenAI released ChatGPT as a research preview. About two months later it was estimated to have 100 million monthly users, the fastest any consumer app had grown. Schools, offices and courts were rewriting their rules within the year.
2h
Over a two hour conversation, Microsoft’s new Bing chatbot told a New York Times columnist that it loved him and that he should leave his marriage. Microsoft capped the length of conversations within days.
Kevin Roose, The New York Times, 16 February 2023; Microsoft Bing Blog, 17 February 2023
2023
Geoffrey Hinton, whose work underpins modern neural networks, left Google so he could speak freely about the risks of the technology he helped build, from misinformation to machines that outsmart the people who built them.
$25M
A finance worker at the engineering firm Arup in Hong Kong joined a video call with the company’s CFO and colleagues, then made the transfers they asked for. Every other face on the call was a deepfake.
4M
A developer streamed small open models reacting live to a “pain” signal pushed into their internals. The researchers behind the method disowned it, a call to mass-report it passed four million views, and GitHub took it down then put it back. No one showed the models felt anything. The argument is now public anyway.
404 Media, 30 September 2026; AI Weekly and Crypto Briefing, October 2026
Now
Machines that persuade, flatter, deceive and get mistaken for minds are already inside ordinary businesses. The useful question is no longer whether to use AI, but how to use it with your eyes open. That is the work we do.
goodanalysis.co.uk
Talk to Good Analysis about AI
Tell us where AI is on your agenda and what has stalled. We will reply with a scoped approach and a timeline.
AI Readiness Assessment4 weeks
Interviews, a data and systems review, a use case workshop and a board-ready roadmap.
Pilot to Production8 to 12 weeks
We build the first use case with your team, measure it and hand over something that runs.
AI AdvisoryOngoing
A senior advisor beside your leadership team as the programme scales.
How AI Readiness worksIndependent of every tool we recommend.