The worst risk when it comes to artificial intelligence is that society would lose control of the technology. That's what Michelle Donelan, Britain's Minister for Research, says, according to The Telegraph.
The statement comes in connection with the world's first major AI summit, which is being organized in the UK over two days. The country has identified three different risks with AI, including that the technology can contribute to the spread of misinformation, but losing control would be the worst, she says.
- It is a risk that is much more hypothetical in nature and naturally worries me the most because it would result in the most serious consequences.
The reporter compares such an outcome to "Terminator", to which Donelan says:
- It could potentially lead to that, but there are several steps before that.
*****************************
Sunak on China-US AI deal: "Good sign"
British Prime Minister Rishi Sunak is defending the decision to invite China to the big AI summit taking place in the UK on Wednesday and Thursday, despite loud criticism from party colleagues. He says this in an interview with Politico's podcast "Power play" before the second day of the AI meeting.
Sunak has gotten nearly 30 meeting participants, including rivals China and the United States, to sign the so-called Bletchley Declaration. Opportunities, risks and future needs around AI are listed there.
That pleases Sunak. He states that it was a long time ago that the USA, Europe and China "wrote on the same paper".
- I think it is a good sign of progress.
******************************
The AI thirst in non-digital languages is creating new jobs
Generative language models such as Chat GPT depend on language data to be able to communicate, which becomes a challenge for the tech giants as soon as they look outside the Western world.
- Over 70 Indian languages spoken by at least one million people have no digital "text corpus". That's how big the problem is, says Manish Gupta at Google's India office.
The solution to the problem could lay the foundation for a job market as hidden as it is gigantic, writes Bloomberg. In India alone, one million people can work with so-called data annotation in 2030, according to the organization Nasscom.
One of the actors is Karya, which both Microsoft and Google rely on. Through the startup's app, tens of thousands of people have been paid to collect and categorize language data. And for that you pay relatively well.
Bloomberg has met Preethi from Agara village. Working for Karya, she reads out sentences in her native Kanarese – a language spoken by over 35 million people – and earns the equivalent of $54 in three days. As a seamstress, she earns less than $1 a day.
- If you pay well, the staff is more invested in the work. The end result is better data, says Saikat Guha of Microsoft Research India.
Inga kommentarer:
Skicka en kommentar