Supremacy is  behind-the scenes story of the ruthless shadow battle between Microsoft and Google and their AI companies, their struggles to use their tech for good and the dangerous direction that they’re now going.

 Human brain power outstripped by AI. Five years ago no one had heard of Open AI. Now the research firm is powering Microsoft’s software, writing everything from code to school essays and bot-distributed misinformation.

In Supremacy AI, ChatGPT and the Race that Will Change the World, journalist Parmy Olson  carefully tracks the rise of two young companies, – the American OpenAI and the British rival DeepMind, following the race to build not just AI models, which includes the kind of technology that helps people to create images or write emails, but AGI, intelligence that could far outstrip human brain power. 

OpenAI the maker of ChatGPT was founded to chase the dream of AGI , even though its chief executive, Sam Altman, has warned of its dangers. Demis Hassabis when he co-founded DeepMind in 2010 alongside  Shane Legg and Mustafa Suleyman, his aim was developing AGI to unpick the mysteries of the universe, from religion to science . Hassabis figured that task would take five years, it’s been 14 years and counting.

OpenAI won early funding from Elon Musk in part because he was suspicious of Hassabis. The world’s richest man saw the British scientist as  a shady character because Hassabis designed a game called Evil Genius .  Olson reports Musk was overheard calling him “the Hitler of AI”.

Both companies believed they needed to reach AGI first – before each other or anyone else.  – as if the only way to ensure the potentially world-changing technology that would be used for good.

AI is expensive. The computer power required is massive tens and thousands of graphics processing units are required to train a system  and each cost £40, 000.

Competition for the best talents  send salaries skyrocketing  the very best are paid millions, that is why large companies now dominate in the creation of AI models.

Best talents were snapped up by Google, Facebook, Apple,  and the like , Open AI and DeepMind sold out to Big Tech. In 2019n AI allowed Microsoft to buy 49 per cent stake to access its billions. DeepMind was bought by Google, although its founders tried to keep the reins by attaching conditions.  DeepMind demanded an ethic board to prevent its yet-to-be-buiilt AGI from being misused. Google allowed the board to meet just once. Olson reports now attempts by Haasabis to ensure oversight were stymied by Google, although it promised more autonomy for DeepMind after its restructuring as Alphabet, that also didn’t happen. Even smart scientists are susceptible to corporate sneakiness.

Microsoft and Google’s money gave them influence and control, but what these two companies wanted was practical AI that could turn profits. So OpenAI must ensure regular updates to its models to keep Microsoft ahead of rivals or put more effort into its image generation tool Dall-E as some bits are freakishly misinformed.  DeepMind has to put its energy into such mundanities as boosting Google’s ad-serving tech and search tools. Hassabis is now Google’s head of AI, with his AGI work reduced to an after-hours hobby.

According to Olson AGI is nothing but a theory – the most transformative technology we’ve ever seen. Olson quotes a former OpenAI manager who says”If you go look at what they’re building it’s just a language model”.

Olson says these mean are not God and Big Tech is not all-powerful . Altman had to backtrack on threats to ditch Europe because of its tighter regulations. Google recently lost a long-running antitrust case over search in the US, and its owner Alphabet faces a potential break-up, California has passed a series of AI laws, and is considering external audit of AI models. AGI could still wreak havoc.

Supremacy AI ChatGPT and the Race That Will Change the World by Parmy Olson, Macmillian Business £22, 336 pages.

One response to “Bot-distributed misinformation”

  1. pennynairprice avatar
    pennynairprice

    AI has only fairly recently reached the ears and eyes of the general public though it has been “simmering” in the background of everything to do with digital enhancement and activities for a while and now is so sophisticated that it can be used to make “cheat pictures”, fake “cheat tapes” (using voice sampling technology and generally producing some information which is “fake” as a result. You can probably find an AI software that will receive a voice message or a set of chapter titles and it can be told to write a book or screenplay in the style of a famous writer? I do beleive this is possible. It can probably compose relatively simple music as well. A must read for anyone in computer technology and the media.

    Like

Leave a comment

Trending