<\/p>\n
Ever since OpenAI , an American firm, launched ChatGPT, its human-like conversationalist, in November 2022, just about every month has brought a flurry of similar news. Against that backdrop, the three latest announcements might look unexceptional. Look closer, though, and they hint at something more profound. The three companies are, in their own distinct ways, vying to become AI national champions. \u201cWe want AI71 to compete globally with the likes of OpenAI”, says Faisal al-Bannai of Abu Dhabi\u2019s Advanced Technology Research Council, the state agency behind the Emirati startup. \u201cBravo to Mistral, that\u2019s French genius,” crowed Emmanuel Macron, the president of France, recently. ChatGPT and other English-first LLMs \u201ccannot capture our culture, language and ethos”, declared Krutrim\u2019s founder, Bhavish Aggarwal. Sarvam started with Indian languages because, in the words of its co-founder, Vivek Raghavan, \u201cWe\u2019re building an Indian company.”<\/p>\n
AI is already at the heart of the intensifying technological contest between America and China. Over the past year they have pledged $40bn-50bn apiece for AI investments. Other countries do not want to be left behind\u2014or stuck with a foreign critical technology over which they have little control. In the past year another six particularly AI-ambitious governments around the world\u2014Britain, France, Germany, India, Saudi Arabia and the United Arab Emirates (UAE)\u2014have promised to bankroll AI to the collective tune of around $40bn. Most of this will go towards purchases of graphics-processing units (GPUs, the type of chips that makes AI intelligent) and factories to make such chips, as well as, to a lesser extent, support for AI firms. The nature and degree of state involvement varies from one wannabe AI superpower to another. It is early days, but the contours of new AI-industrial complexes are emerging.<\/p>\n
Start with America, whose tech firms give everyone else AI envy. Its vibrant private sector is innovating furiously without direct support from Uncle Sam. Instead, the federal government is spending around $50bn over five years to increase domestic chipmaking capacity. The idea is to reduce America\u2019s reliance on Taiwanese semiconductor manufacturers such as TSMC, the world\u2019s biggest and most sophisticated such company. Supplies from Taiwan could, fear security hawks in Washington, be imperilled should China decide to invade the island, which it considers part of its territory.<\/p>\n
Another way America intends to stay ahead of the pack is by nobbling rivals. President Joe Biden\u2019s administration has enacted brutal export controls that ban the sale of cutting-edge AI technology, including chips and chipmaking equipment, to adversaries such as China and Russia. It has also barred Americans from sharing their AI expertise with those countries.<\/p>\n
It is now coercing those on the geopolitical fence to fall in line. In October the American government started requiring companies in third countries, including Saudi Arabia and the UAE, to secure a licence in order to buy AI chips from Nvidia, an American company that sells most of them. The rules have a \u201cpresumption of approval”. That means the government will \u201cprobably allow” sales to such firms, says Gregory Allen, who used to work on AI policy at the Department of Defence\u2014as long, that is, as they do not have close ties to China. On December 6th Peng Xiao, who runs a state-backed AI startup in Abu Dhabi called G42, announced that the company would be cutting ties with Chinese hardware suppliers like Huawei, a Chinese electronics company. \u201cWe cannot work with both sides,” he told the Financial Times.<\/p>\n
China\u2019s AI strategy is in large part a response to American techno-containment. According to data from JW Insights, a research firm, between 2021 and 2022 the Chinese state has spent nearly $300bn to recreate the chip supply chain (for AI and other semiconductors) at home, where it would be immune from Western sanctions. A lot of that money is probably wasted. But it almost certainly helped Huawei and SMIC, China\u2019s biggest chipmaker, design and manufacture a surprisingly sophisticated GPU last year.<\/p>\n
The central and local authorities also channel capital into AI firms via state-backed \u201cguidance funds”, nearly 2,000 of which around the country invest in all manner of technologies deemed to be strategically important. The Communist Party is guiding private money, too, towards its technological priorities. Often it does so by cracking down on certain sectors\u2014most recently, in December, video-gaming\u2014while dropping heavy hints about which industries investors should be eyeing instead. The government is also promoting data exchanges, where businesses can trade commercial data on everything from sales to production, allowing small firms with AI ambitions to compete where previously only large data-rich firms could. There are now 50 such exchanges in China.<\/p>\n
Elements of this state-led approach are now being emulated in other parts of the world, most notably in the Gulf\u2019s petrostates. Being autocracies, Saudi Arabia and the UAE can move faster than democratic governments, which must heed voters\u2019 concerns about AI\u2019s impact on things like privacy and jobs. Being wealthy, they afford to buy both the necessary GPUs (on which the two countries have together so far splurged around $500m) and the energy needed to run the power-hungry chips.<\/p>\n
They can also plough money into developing human capital. Their richly endowed universities are quickly climbing up global rankings. The AI programme at King Abdullah University of Science and Technology, a Saudi institution, and the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) in Abu Dhabi, the world\u2019s first AI-focused school, have poached star professors from illustrious institutions such as University of California, Berkeley, and Carnegie Mellon University in Pittsburgh. And nearly all of MBZUAI\u2019s graduates, who number a couple of hundred, stay in the region to work at local firms and labs, says its provost, Timothy Baldwin (himself lured to the Middle East from the University of Melbourne).<\/p>\n
The Gulf approach is producing results. The capabilities of the Falcon model, first built by a team of 20 or so engineers, rival those of Llama 2, the most widely used \u201copen-source” model devised by Meta, an American tech giant. AI71 plans to improve its open-source models using national data sets from fields including health, education and, some day, perhaps oil. \u201cIn the last 50 years, oil drove the country\u2026now data is the new oil,” says Mr al-Bannai.<\/p>\n
A third group of governments is combining elements of America\u2019s approach with those of the Chinese and Emiratis. The EU has its version of America\u2019s incentives for domestic chipmaking. So do some member states: Germany is footing a third of the \u20ac30bn ($33bn) bill for a new chip factory to be built there by Intel, an American chipmaker. Outside the bloc, Britain has promised to funnel \u00a31bn ($1.3bn) over five years to AI and supercomputing (albeit without going into detail about how exactly the money will be spent). India\u2019s government is promoting manufacturing, including of semiconductors, with generous \u201cproduction-linked incentives”, encouraging big cloud-computing providers to build more Indian data centres, where AI models are trained, and thinking about buying $1.2bn-worth of GPUs.<\/p>\n
Like China and the Gulf but unlike America, where federal and state governments are reluctant to part with public data, India and some European countries are keener on making such data available to companies. France\u2019s government \u201chas been very supportive” in that regard, says Arthur Mensch, Mistral\u2019s boss. Britain\u2019s is considering allowing firms to tap rich data belonging to the National Health Service. India\u2019s government has enormous amounts of data from its array of digital public services, known as the \u201cIndia Stack”. Insiders expect it eventually to integrate Indian AI models into those digital services.<\/p>\n
In contrast to China, which regulates consumer-facing AI with a heavy hand, at least for now Britain, France, Germany and India favour light-touch rules for AI or, in India\u2019s case, none at all. The French and German governments have soured on the EU\u2019s AI Act, the final details of which are currently being hotly debated in Brussels\u2014no doubt because it could undermine Mistral and Aleph Alpha, Germany\u2019s most successful model-builder, which raised \u20ac460m in November.<\/p>\n
It is natural for countries to want some control over what may prove to be a transformational technology. Especially in sensitive and highly regulated sectors such as defence, banking or health care, many governments would rather not rely on imported AI. Yet each flavour of AI nationalism also carries risk.<\/p>\n
America\u2019s beggar-thy-neighbour approach is likely to antagonise not just its adversaries but also some of its allies. China\u2019s heavy regulation may offset some of the potential gains from its heavy spending. Building models for local languages, as Krutrim and Sarvam in India plan to do, may prove futile if foreign models continue to improve their multilingual capabilities.<\/p>\n
The Gulf\u2019s bet on open-source models may misfire if other governments limit their use, as Mr Biden has hinted at in a recent executive order and the EU could do through its AI Act, out of fear that open LLMs could too easily get into the hands of mischief-makers. Saudi and Emirati institutions may struggle to hold on to talent; a developer who worked on Falcon admits it benefited greatly from a partnership with a French team of engineers who have since been poached by Hugging Face, a high-flying Silicon Valley AI startup. As one sceptical investor notes, it is not yet clear how vast or useful public Emirati data actually is.<\/p>\n
Handing companies sensitive data on things like citizens\u2019 health could spark a public backlash even in autocratic places, let alone Britain, France or Germany. As for industrial policy, it has a lousy record of spurring innovation and economic growth when the industry in question is mature, which AI is not. Picking winners in a fast-changing field verges on the foolhardy.<\/p>\n
As Nathan Benaich of Air Street Capital, a venture-capital firm, sums it up, most efforts to create national models \u201care probably a waste of money”. This warning will not dissuade AI-curious governments, mindful of the rewards should they succeed, from meddling. Mr Macron will not be the only leader to greet it with a Gallic shrug.<\/p>\n<\/p><\/div>\n","protected":false},"excerpt":{"rendered":"
Ever since OpenAI , an American firm, launched ChatGPT, its human-like conversationalist, in November 2022, just about every month has brought a flurry of similar news. Against that backdrop, the three latest announcements might look unexceptional. Look closer, though, and they hint at something more profound. The three companies are, in their own distinct ways, […]<\/p>\n","protected":false},"author":2,"featured_media":222145,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[7],"tags":[440,263,262,5834,260,259,258,3686,265,202,5792,261,11979,15804,264],"class_list":["post-222144","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-metaverse","tag-ai","tag-axie-infinity","tag-axs","tag-chatgpt","tag-decentraland","tag-facebook","tag-game","tag-huawei","tag-mark-zuckerberg","tag-nft","tag-openai","tag-sandbox","tag-smic","tag-technological-destinies","tag-vr"],"_links":{"self":[{"href":"https:\/\/dripp.zone\/news\/wp-json\/wp\/v2\/posts\/222144","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dripp.zone\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dripp.zone\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dripp.zone\/news\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/dripp.zone\/news\/wp-json\/wp\/v2\/comments?post=222144"}],"version-history":[{"count":2,"href":"https:\/\/dripp.zone\/news\/wp-json\/wp\/v2\/posts\/222144\/revisions"}],"predecessor-version":[{"id":222151,"href":"https:\/\/dripp.zone\/news\/wp-json\/wp\/v2\/posts\/222144\/revisions\/222151"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dripp.zone\/news\/wp-json\/wp\/v2\/media\/222145"}],"wp:attachment":[{"href":"https:\/\/dripp.zone\/news\/wp-json\/wp\/v2\/media?parent=222144"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dripp.zone\/news\/wp-json\/wp\/v2\/categories?post=222144"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dripp.zone\/news\/wp-json\/wp\/v2\/tags?post=222144"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}