Meta Platforms, which also owns WhatsApp and Threads, a rival to Twitter/X, reported that revenues rose 25 per cent to $40.1 billion for the three months to the end of December. ­Analysts were expecting revenues of $39.2 billion.

The company announced a dividend of 50 cents a share and an additional $50 billion share buyback as profit rose to $10.6 billion in the quarter from $300 million a year earlier.

The technology giant forecast first-quarter revenues of $34.5 billion to $37 billion, above Wall Street expectations of $33.8 billion. It ­expected total expenses for this year to be unchanged at $94 billion to $99 billion.

Meta said that advertising impressions, or views, increased 21 per cent from a year earlier and that the average price per advert rose by 2 per cent.

Facebook reported 2.11 billion daily active users on average for December, an increase of 6 per cent year on year, while monthly active users stood at 3.07 billion — a rise of 3 per cent.

This week Alphabet, the owner of Google, a ­fellow digital advertising heavyweight, posted ­results that disappointed Wall Street, after holiday season sales were below expectations.

“This was one of the most impressive quarters intrinsically and versus expectations,” Mark Mahaney, an analyst at Evercore, an investment bank advisory firm, said of Meta’s results.

Shares in the company, which is based in California and was founded in 2004, rose by $55.88, or 14.1 per cent, to $450.78 in late trading, lifting its value to $1 trillion.

Improvements to the social media business have made shareholders more tolerant of Meta’s undiminished spending, with huge investments in “metaverse” technologies and its artificial intelligence infrastructure. Its recovery has been aided by a rebound in user growth and digital ad sales. It has also more than 21,000 employees since late 2022.

“Meta ended 2023 on an extremely strong note, with revenue soaring above analyst expectations,” Debra Aho Williamson, a tech analyst at eMarketer, said. “The company can talk all it wants to about artificial intelligence and the metaverse, but it’s still a social media company that gets nearly all its revenue from advertising, and ­advertisers still clearly love Meta.”

The company’s metaverse-oriented Reality Labs division exceeded revenue expectations for the fourth quarter, posting record sales of $1.1 billion from “strong sales” of its Quest device over the holiday season, Mark Zuckerberg, chief executive and co-founder of the company, said after the report. Investors had been expecting $804 million.

“We’ve made a lot of progress on our vision for advancing AI and the metaverse,” Zuckerberg said.Meta said that it still expected operating losses for Reality Labs to “increase meaningfully” as it invests further in augmented and virtual reality.

The booming revenues at ByteDance, up from $80 billion last year, show how the social app is booming in popularity around the world, although it is not thought to be profitable.

According to the Financial Times, ByteDance’s American business made up more than 10 per cent of the total, a record $16 billion of sales. TikTok would not comment on the figure. ByteDance is a privately owned company and so its financial information is not made public.

TikTok has long been viewed as controversial by western governments because of its Chinese founders and its enormous presence in the country. With its addictive algorithm and mass appeal, there are fears about the misuse of people’s data, as well as the threat of undue influence over users by the Chinese government.

As a result, a ban looms over the Chinese social media business in the United States. On Wednesday, the House of Representatives voted through a bill that would force ByteDance to sell TikTok to a new owner within six months or face being wiped off American app stores, such as Apple and Google. The bill has to go to the Senate for another vote and then it must be signed off by President Biden in order to be made into law.

Donald Trump tried to ban TikTok in 2020 when he was president, but the move was overturned by the courts. It is banned in India.

“TikTok is too large and valuable for the platform not to be sold to one of many potential buyers,” analysts at Jefferies said, noting that there was “a less than 10 per cent chance” that the service would be totally shut down in America. Dan Ives, an analyst at Wedbush Securities, put the likelihood higher, at 25 per cent.

TikTok has an estimated 170 million users in the US and about 150 million throughout Europe. The latest annual report from Meta, which also owns Instagram, Threads and WhatsApp, said it had 2.11 billion daily active users. Snap reported in its most recent earnings statement that it had 414 million daily active users and had made $4.6 billion in annual revenue, almost flat on the year before.

Those who have expressed an interest in taking over the app include Steve Mnuchin, the former US Treasury secretary, who said he was pulling together a group of investors for a bid.

Another prominent figure said to be looking at the business is Bobby Kotick, the founder of Activision Blizzard, the computer games business behind Call of Duty. He has stepped back from his role at Activision since its acquisition by Microsoft. During previous discussions about a ban in 2020, Microsoft also expressed an interest.

TikTok could be worth between $35 billion and $40 billion, according to Bloomberg Intelligence.

ByteDance maintains that while the company has Chinese founders, it is not a Chinese business, with 60 per cent owned by investors such as The Carlyle Group and General Atlantic, 20 per cent by its co-founder and 20 per cent by its workforce. Three out of five of its board members are American.

It says that it does not hold data in China, nor does it share it with Beijing and that it is spending €12 billion on “Project Clover” to protect European and British users’ data, with a similar “Project Texas” in the US. Since 2022, Oracle, the American technology company, has held the data of US users on its servers.

Mercedes, Tesla and Volkswagen are among the manufacturers scrapping many conventional hands-on controls. Instead, drivers have to interact with a touchscreen, typically sited to the left of the steering wheel, which means they must look away from the road.

Under the rules, to be introduced in January 2026, any car seeking maximum points for the highest safety rating of five stars must use buttons, stalks or dials for five critical tasks: indicating directions, triggering hazard lights, sounding the horn, operating windscreen wipers and activating the eCall SOS function, which automatically calls the emergency services in the event of a serious collision.

The intervention by Euro NCAP, the automotive industry safety body, is a warning shot to carmakers that are rushing to install ever more prominent touchscreens on their vehicles’ dashboards. The Mercedes EQS, for example, boasts a 56in “hyperscreen”.

Screens were initially used for sat navs and infotainment, such as music and radio playback, but the trend has been to keep adding driving functions traditionally accessed through physical controls.

It is tempting for carmakers to eliminate hands-on controls where possible because it not only results in a sleek and minimalist look for the cabin but saves on production costs, reduces weight and enables the manufacturer to easily update the software.

But safety campaigners are dismayed at the trend, which they fear will undermine compliance with the ban on using mobile phones at the wheel. The offence carries a punishment of six penalty points and a £200 fine, alongside the loss of a licence for anyone who only passed their driving test in the previous two years.

Matthew Avery, director of strategic development at Euro NCAP, the car safety body, said: “The overuse of touchscreens is an industry-wide problem, with almost every vehicle-maker moving key controls onto central touchscreens, obliging drivers to take their eyes off the road and raising the risk of distraction crashes.

“New Euro NCAP tests due in 2026 will encourage manufacturers to use separate, physical controls for basic functions in an intuitive manner, limiting eyes-off-road time and therefore promoting safer driving.”

Although the safety rules are not mandatory, carmakers put great store on winning a five-star safety rating.

The latest Tesla Model 3 presents the driver with a blank dashboard behind the steering wheel. The motorist must instead look to the centre console to check their speed and tap buttons using arrows on the steering wheel to indicate a turn to the right or left, even if the steering wheel is moving.

The new Volkswagen iD7 dispenses with knobs to control the air vents, which can only be accessed via a screen-based climate control menu, while the Volvo EX30 requires the driver to go into a sub-menu to adjust the electric mirrors and the rear window de-mister. By contrast, Mazda has opted to keep conventional controls on its cars.

Edmund King, president of the AA, who tested the latest Tesla Model 3 last week, wrote in a blog: “I’ve tried not to be a luddite, but I did find the total lack of any dashboard dials very daunting. In front of the steering wheel, there is a void. There are no dials. Getting into the car late at night and pulling away felt like I had no lights on, as nothing illuminates the dash in front of you — because there isn’t one.”

He said: “Car [controls] need to be intuitive for all drivers so they can concentrate on the road ahead rather than the screen to the side. Euro NCAP has been transformative in crash protection and saved countless lives, so it is reassuring that it is addressing how design may lead to driver distraction.”

Laws will soon come in across Europe enforcing mandatory in-cabin monitoring to detect whether the driver is distracted. An alarm will sound if the driver is distracted for more than 3½ seconds at speeds of 30mph and above, and for six seconds below 30mph. They are set to be followed for cars sold in Britain.

Cameras in the car will analyse the driver’s eye and face movements to determine their attention levels; some models will use AI to that end.

The new regulations will require all new car types to be equipped with advanced driver distraction warning systems from July 7 and this will be extended to all new registered vehicles from July 7, 2026.

A report published in January by the European Road Safety Observatory found that drivers engaged in distracting activities for about half of all driving time, with the main causes being using a mobile phone, adjusting infotainment systems, interacting with passengers and eating.

Mike Hawes, chief executive of the SMMT, the auto industry trade body, said: “Safety is the number-one priority for the car industry. While manufacturers will have their own individual strategies for designing in-car controls, the latest systems enable the driver to dedicate maximum attention to the road ahead while ensuring interaction is as straightforward as possible.”

He said that voice-activation can complement physical buttons and screens, with some functions — such as sending a text message — being operated by voice command.

“Come here, Green,” he said. “Stop wasting time.” Green didn’t move. Orange, an identical robot that had also ambled on stage, was more co-operative. It clomped over and stopped at Huang’s command.

Green’s apparent intransigence was a small fly in the ointment at the showcase conference of Huang’s $2.2 trillion chip giant. But his enthusiasm for this fantastical future was undimmed. It is a future where bipedal, humanoid robots are moments away from bursting onto the world, powered by a novel robotic operating system Nvidia has developed. “Everything that moves in the future will be robotic,” he enthused. “It won’t just be you.” Indeed, before Green and Orange showed up, Huang strode on stage in front of a row of stationary humanoid robots, like a drill sergeant surveying his new recruits.

Since OpenAI’s launch of ChatGPT, the chatbot that wowed the world 18 months ago, investors have ploughed an estimated $100 billion into artificial intelligence (AI) start-ups. The AI age, it appears, has begun. But while the world has been transfixed by chatbots and tools that transform text prompts into jaw-dropping videos, a frenzy over robots powered by AI “brains” has taken hold that insiders claim is about to shock the world all over again.

“There are more than 30 companies that I’m aware of that are developing humanoid robots with plans to commercialise in the next year,” Andra Keay, an investor and managing director of trade group Silicon Valley Robotics. She compared their arrival to other world-altering inventions: “Humanoid robots are the automobiles of the 21st century.”

Doubters will guffaw at the idea that droids will soon flood into factories, fast-food kitchens and our homes. Two years ago, Elon Musk announced that Optimus, the humanoid robot being developed by Tesla, would be, “bigger than the car business”. In the event, Optimus ended up being a human dressed in a robot bodysuit who danced on stage.

Green’s on-stage glitch last week could easily be painted as yet another example of why robots are still a very long way from being useful. And yet, a growing number of investors are making huge bets that leaps in AI systems and plummeting component costs are converging with an acute labour shortage and rising wages for human workers to create the ideal conditions for the robots to finally arrive.

Figure AI, a two-year-old Sunnyvale, California start-up, raised $675m this month to commercialise its bipedal bot. Founder Brett Adcock, who previously set up flying-car company Archer Aviation, predicted that human labour would soon be automated away entirely. “We are in the early stages of an AI and robotics revolution. Everywhere from factories to farmland, the cost of labour will decrease until it becomes equivalent to the price of renting a robot, facilitating a long-term, holistic reduction in costs. Over time, humans could leave the loop altogether as robots become capable of building other robots,” he said. “We have the potential to alter the course of history and fundamentally improve millions of lives.”

For years, companies have had to painstakingly train their robots to carry out discrete tasks such as picking up an object and setting it on a table. Ask that robot to, say, grab a hairbrush, or walk across a warehouse, and they are useless. Breakthroughs in AI, however, have led to AI-based reasoning models that are better at understanding the world around them, comprehending what is asked of them, and responding accordingly.

Pras Velagapudi, chief architect of Agility Robotics, a developer of a bipedal warehouse robot called Digit, said: “The newest AI models are very promising in solving some of the upcoming problems that we had anticipated would take years or decades to solve. We’re going to see these types of robots out in the world much faster than we originally anticipated.”

Predictions of mass automation have been made countless times, only to fizzle. And there is something almost cartoonish about robots that are made to look and move like humans. It calls to mind the famed quote about the automobile from Henry Ford: “If I had asked people what they wanted, they would have said faster horses.”

But there are good reasons to try to replicate humans. American warehouse and manufacturing companies have more than 600,000 unfilled positions. Despite the need, many have been reticent about automation because it often requires remaking an operation from the ground up to make an environment suitable for even a modicum of additional automated work.

Therein lies the allure of humanoid robots. Because they are facsimiles of humans in size, shape and locomotion, no complex retrofit is required. A capable droid can, theoretically, be dropped into an existing operation.

A cursory scan of YouTube for demo videos might lead one to think that the robot age is already here. Figure posted a demo of its bot, powered by OpenAI’s language model, that displayed an impressive ability to understand the world around it. When asked for something to eat, it handed its interlocutor an apple. It also appeared dextrous, complying with voice commands to pick up rubbish and stack (plastic) dishes in a drying rack. Tesla published a video in December of the latest iteration of Optimus, which could walk 30 per cent faster than the previous model. China’s Unitree has shown its bots running and climbing up and down stairs.

Dwight Klappich, an analyst at tech consultancy Gartner, is more circumspect. The path from a whizzy video to a truly capable robot capable of operating in the real world, and all the variables it presents, is very long indeed. “Over the next 20 years, you’ll go into a McDonald’s and it’s going to look like Star Wars. There will be robots doing a lot of the work.”

Huang, Musk and the rest are betting that future will arrive far faster.

On the anniversary of the collapse of Silicon Valley Bank, the IMF has rung the alarm bell over the risks of another round of bank failures triggered by the worst fall in commercial property values in half a century in the world’s largest economy.

“The high concentration of corporate real estate exposures represents a serious risk to small and large banks amid economic uncertainty and higher interest rates, potentially declining property values, and asset quality deterioration,” the IMF said in an analysis of financial stability risks a year on from SVB’s failure.

Rising US interest rates have caused an 11 per cent fall in US commercial property prices since 2022 which are on course for another 10 per cent decline this year. Small and regional American banks lend heavily to the sector, with about a third of the entire US banking system heavily exposed to balance sheet losses if loans turn sour, the IMF said.

Financial stability concerns have returned to parts of the banking system in recent weeks after the regional lender New York Community Bancorp lost a quarter of its share price value after reporting losses tied up to its exposure to corporate real estate. Banks are having to set aside larger amounts of cash to protect themselves against bad loans, with lenders “anticipating additional defaults”, the IMF said.

“There are reasons to expect non-performing loans to climb further in the coming quarters, for example in the United States, quarterly corporate real estate non-performing loans and losses did not peak until nine quarters after the start of the global financial crisis in mid-2007,” the fund said.

In March 2023, SVB and a host of US regional banks went bust over their bets on US government bonds, whose prices have slumped in an era of tight monetary policy. SVB also suffered from mass deposit withdrawal from its customers, who were concentrated in the west coast tech sector.

A year after the financial tremors, the IMF said a “sizeable subgroup” of banks still faced difficulties in managing their balance sheets and retaining deposits.

“Underlying concerns persist, with fears that the failure of one institution could precipitate a broader loss of confidence in the sector. Beyond the unrealised losses due to higher interest, the credit risk carried by some institutions, particularly their exposure to corporate real estate, is at the centre stage of investors’ fears today,” the IMF said.

The IMF and institutions like the Bank for International Settlements have also highlighted the dangers of recent market volatility after investors have rapidly repriced their bets on interest rate cuts and are expecting fewer rate cuts from the US, UK, and eurozone central banks this year. Government bond prices have fallen further in recent months as a result.

The SVB crisis prompted US regulators to offer a state guarantee against all the bank’s deposits in an attempt to stem contagion. The UK subsidiary of SVB was spared from collapse after an eleventh-hour rescue by HSBC, which bought the lender’s assets for £1, ensuring that British taxpayers were not on the hook.

Now, assuming that you are on the right side of 50, your first reaction might be “What’s Amstrad?” Well, pull up a chair: a long time ago the UK had a home computer industry and Amstrad was part of it. The company, which started in low-priced consumer electronics, was quite big in home computing in the Eighties, but foundered in the Nineties. It produced a strange landline phone that sent emails in 2000. Finally, it was taken over by BSkyB in 2007 (it was a big supplier of set-top boxes).

Amstrad may be a footnote in technology history, but its founder still looms in British media. He is best known for fronting The Apprentice in the UK. Sugar said that he had wanted to buy back the name (which is derived from “Alan Michael Sugar Trading”) for years, but Sky’s “belligerent” management had refused. In 2018, however, Sky became a subsidiary of Comcast, paving the way for the rebirth of Amstrad.

All very interesting, but I still find myself thinking: “Amstrad? Amstrad?” I am old enough to remember the glory days of home computing and I smile when I recall brands such as Acorn, Commodore, Texas Instruments and Atari. But Amstrad? Amstrad? As the kids say, I’m just not feeling it. Until last week’s news, the brand had not crossed my mind for 30 years.

Wary of generalising from my own experience, I rang round a few contacts who work in this area. Hannah Williams, a tech PR said: “If you do remember it, you probably don’t remember it fondly.”

Another, a friend who works in branding (and who is closer in age to my daughters than me) asked: “What’s Amstrad? It sounds like a German transport interchange.”

I have always been interested in rebranding, relaunches and refreshes, not least because millions of pounds are often spent on very public car crashes. Even in the world of branding, though, this is something pretty unusual. It is full-on brand resurrection, a name raised from the dead. And this does not happen often. You sometimes see brands disappear for a few years. But almost two decades?

To be fair, there is a cottage industry in reviving long-forgotten brands. Years ago I interviewed a guy who specialised in drinks brands. He would send researchers into the cellars of places such as Fortnum & Mason looking for interesting old booze names that had disappeared decades or even centuries ago and, although it is true that a long-dead gin brand will not have much in the way of public awareness, it’s likely to have a good story and the right feel. I don’t think, though, that Amstrad is quite the same as an art deco name on a charming old bottle.

Scott Perry, a brand strategy consultant, says that brands gain their emotive power by engaging with their audience over time and have to work hard to maintain this. “But if you leave a brand without engagement, the equity you’ve built up is likely to depreciate. The world moves on fast.”

Sugar clearly is untroubled by this. He said: “They’ve got the brand name: that in itself is worth a fortune and they’ve got to exploit it. And I want to see some money.”

For what it’s worth, I get the attachment that people feel to brands that they have created. A recent example is Angie Louise 3 Monkeys Communications Moxham. The story is here is that she sold her PR firm, 3 Monkeys Communications, to Zeno Group in 2016. Zeno, according to PR Week, retired the name in 2023. Moxham now wants it back, but Zeno owns it and, she says, will not let her use it. So Moxham has legally changed her name in the hope of reinforcing “the idea that theagency name is synonymous with her own personal brand”.

As a friend who has advised on a number of branding spats put it: “The point surely comes when it’s easier just to make up something new.”

Here, we can learn from the Dassler brothers who were behind two of the world’s best-known sports bands. They initially ran a company called Gebrüder Dassler Schuhfabrik, which translates as “Dassler Brothers Shoe Factory” and was shortened to Geda. Shortly after the Second World War they fell out. As a result, they set up two new brands: Adolf “Adi” Dassler founded Adidas and Rudolf formed Ruda, which would later become Puma. How sensible.

In a similar vein, Jeremy King’s new reboot of Princess Diana’s favourite restaurant, Le Caprice, goes by another name. It is on the same site and will draw on its predecessor’s heritage, but it is called Arlington because King does not own the name (and it is on Arlington Street in St James’s).

Perhaps brand-fixation is down to the modern business obsession with IP. The idea is that because something has existed previously in the public consciousness, it must have some sort of talismanic worth. Although, as Perry said: “The value in a brand is something other people see in it, not what you see in it.” He added that the desire to resurrect long-dormant brands that have meaning for you but little public resonance is about romance and nostalgia, not brand value.

Of course, Sugar is a billionaire and can do as he likes. And who knows? Perhaps the brand will become a household name again. And, if it does, I’ll be forced to eat any Amstrad products that may be gathering dust in my parents’ loft. Perhaps Amstrad Digital will even deliver the money Sugar wants to see. But even so, I can’t stop myself thinking: “Amstrad? Amstrad?”

Rhymer Rigby is a journalist and author. Follow him on Twitter @rhymerrigby

The veteran Silicon Valley investor Marc Andreessen simply posted “what” on Twitter/X. Even Jeremy Hunt, the chancellor, had a view: “Quite a thing to see the top roles in AI, both at DeepMind/Google and Microsoft go to extraordinary British talent — helping us realise our ambition to be the world’s next Silicon Valley.”

In a nod to the commercial nous of Satya Nadella, Microsoft’s energetic chief executive, one X user remarked “Satya is the godfather”, next to a picture of Robert De Niro with the caption: “I’ll make you an offer you can’t refuse.”

Perhaps the most perceptive comment came from another user who said: “How come u don’t call it mAIcrosoft?” A rebrand that would certainly encapsulate how the tech group is jumping into the developing technology with both feet.

Suleyman’s defection is being hailed as a coup for Microsoft and yet another AI chess move from Nadella, who is systematically gathering up some of the world’s top artificial intelligence talent from making investments and alliances with start-ups and people.

As Brad Smith, Microsoft’s president, said at this year’s Mobile World Congress: “We need to advance a broad array of AI partnerships. Innovation and competition will require an extensive array of similar support for proprietary and open-source AI models, large and small. We have also invested in a broad range of other diverse generative AI startups. In some instances, those investments have provided seed funding to finance day-to-day operations. In other instances, those investments have been more focused on paying the expenses for the use of the computational infrastructure needed to train and deploy generative AI models and applications.”

Far from the boring creator of Excel spreadsheets and Word, Microsoft is trying to stay one step ahead of the competition by boldly going into the world of AI, making sure that every corner of the tech is linked back to the mothership. The concept of “mAIcrosoft” feels spot on.

Microsoft’s AI advances helped to lift the group’s second-quarter revenues to $62 billion. The group has been rolling out Copilot, an AI tool for businesses embedded in its popular software products, that can draft emails, make presentations and send out meeting transcripts.

Early sales were reflected in the company’s commercial sales of Office software, where revenue grew by 17 per cent. Analysts at Wedbush have said that the technology could increase Microsoft’s revenues by about $25 billion.

What does this mean for Inflection AI, which raised more than $1.3 billion in 2023, including from Microsoft? Only a week ago it launched Inflection 2.5, a close competitor to OpenAI’s latest ChatGPT model, and reported having one million daily active users.

The investors “will have a good outcome today and I anticipate good future upside”, according to Reid Hoffman, Inflection AI’s co-founder and the founder of LinkedIn, which was bought by Microsoft in 2016. Hoffman also sits on Microsoft’s board.

Microsoft has paid Inflection $650 million, mainly for a licence fee to make its models available on its Azure cloud service.

Tony Wang, managing partner at 500 Global, a venture capital company with $2.4 billion under management, said: “My initial reaction was shock … the fact that Mustafa left his company on a rising trajectory to join a competitor incumbent broke all models of pattern matching.” He called it “an acquisition of Inflection without having to go through regulatory approval”.

Many regard it as an “acqui-hire”; hiring a talented bunch of employees as a way of buying a business without some of the hassles, not to mention regulatory scrutiny, of an acquisition.

Nonetheless, Inflection lives on without them. “@InflectionAI will continue on its mission under a new CEO, and look to reach more people than ever by making its API widely available to developers and businesses the world over,” Suleyman said.

Microsoft’s closest relationship within the AI start-up scene is with the ChatGPT maker OpenAI. It has poured $13 billion into the business, receiving an undisclosed profit share, thought to be 49 per cent. In return, it gives OpenAI access to its vast cloud infrastructure, the power it needs to train its compute-hungry large language models that sit behind its chatbot.

Something similar to this “acqui-hire” almost happened between Microsoft and OpenAI last year, when Sam Altman was pushed out of the business and then brought back over the course of a fraught weekend. Nadella offered him a role within Microsoft and offered to take as many of the team with him as wanted to come.

Altman, OpenAI’s chief executive and co-founder, was by Nadella’s side throughout this year’s Davos conference as the two were feted for blazing a trail in AI, bringing it to the consumer market through Microsoft’s products.

Some believe this kind of manoeuvre from Microsoft was a takeover in all but name.

All these frenzied links between Big Tech and the AI world have attracted the attention of regulators. In the US, the Federal Trade Commission (FTC) has called on Alphabet, Amazon and Microsoft to supply information on their investments and partnerships with OpenAI and its rival Anthropic, backed by Google and Amazon.

In the UK, the competition authority is also investigating whether the relationship between Microsoft and OpenAI could be considered a merger. The European Commission is looking into whether Microsoft’s multibillion-dollar investment in OpenAI represents a breach of its antitrust rules.

The Inflection deal may not just be about people, it could also be about its chips or computing power.

According to the latest annual State of AI report, Inflection owns 22,000 of Nvidia’s powerful H100 chips, more than Oracle or Tesla, which the report said have 16,000 and 10,000 respectively as of last month. In 2023 Inflection announced that it was “building one of the largest computing clusters in the world” as part of a tie-up with CoreWeave and Nvidia.

These will stay with the start-up for the moment, although it is not clear what will happen to them in future. A spokesman for Inflection said: “Inflection has a contract with CoreWeave for the cluster and that contract is still in place. At some point in the future it may or may not be amended, which may or may not release the cluster to be used by someone else. But that’s not a decision that’s been taken.”

In the global arms race for the compute that powers AI, these much-in-demand graphics processing units are highly valued. They are “harder to buy than drugs”, as Elon Musk quipped last year.

Even though he was no longer at Google DeepMind when he made the jump, Microsoft bringing in Suleyman, a former member of its rival’s AI business, is symbolic of the intensity of the AI arms race at the heart of Silicon Valley.

Liz Kendall, the work and pensions secretary, says that “spiralling inactivity is the greatest employment challenge for a generation” and that reversing it is central to Labour plans to boost economic growth.

There are now a record 2.8 million people off work due to long-term sickness, up 700,000 since Covid-19, with the cost of sickness benefits due to reach £64 billion by the end of the parliament, up £30 billion on before the pandemic.

A report by the NHS Confederation and the Boston Consulting Group found that reversing even part of the impact would save taxpayers £19.5 billion a year by the end of the decade through lower benefit spending and higher tax receipts. The boost to the wider economy would be even bigger, at £62.6 billion.

The report also highlights that Britain has become an international outlier, with sickness up by 27 per cent since the pandemic while it has stayed the same or fallen in the rest of Europe. In Britain one in every 15 people of working age is now off owing to long-term illness, a rate 69 per cent higher than Germany and more than twice Italy’s level.

Analysis by BCG of people’s routes into long-term sickness found that students were now one of the biggest contributors. In 2021-22, 63,392 people went straight from being economically inactive because they were studying to being inactive through long-term sickness, up from 36,866 in 2019-20. Once people flowing the other way are subtracted, students pushed up inactivity numbers by 42,300 net, up from 12,700 in 2019-20.

While the rate had fluctuated before the pandemic, the figures show a 24 per cent rise since 2014 and an average annual net increase of 5.5 per cent, with rising numbers of students flowing into long-sickness while those moving from sickness to study remain more constant.

• NHS offers job coaching for patients with depression and anxiety

Raoul Ruparel, author of the report, said: “Those aged 16-24 who are long-term sick inactive with multiple conditions has grown by 18 per cent since pre-Covid, making them the fastest growing group of those out of the workforce. This is largely driven by an acceleration of mental health conditions post-Covid. It is no coincidence then that the numbers of those moving directly from being students to being economically inactive has grown rapidly — these flows have more than tripled since pre-Covid.”

He said the growing trend “paints a concerning picture. Not only is this meant to be the healthiest section of the population but those who spend a prolonged period out of work at the start of their career have been shown to have materially lower income and productivity for the rest of their lives.”

His report called for earlier mental health support and better help from universities and employers. Matthew Taylor, head of the NHS Confederation, said the health service could not solve the long-term sickness problem alone.

• David Smith: How can we get 80% of Britons into work? What history tells us

“The UK faces a series of enormous health challenges as seen in the sharp rise in the number of people out of the workforce due to long-term sickness. This spike has defied European trends and requires both additional investment in the NHS alongside coordinated and sustained action across government,” Taylor said.

“Getting this right will lead to a more productive NHS, support more people to be in work and provide a major boost to the economy. Failure risks rising waiting lists and the proportion of public spending spent on the NHS crowding out other forms of investment, which will then only weigh down the economy.”

On Monday Kendall held the first meeting of a labour market advisory board as she plans a white paper on tackling economic inactivity expected after the budget. It is expected to focus on devolving powers to local mayors to join up employment, health and skills support services.

She said that dealing with the problem “will take time, but we’re going to fix the foundations of the economy and tackle economic inactivity”.

Gemini is its version of an LLM, or large language model, the technology that underpins chatbots such as ChatGPT. Google claims it’s the first AI model to do better than humans on a popular knowledge and problem-solving test. You can use it now through Bard, its AI chatbot, provided that you’re in one of the 170 countries where it’s available (it’s not yet available in the UK, but is coming soon).

The implication is that Google’s product is the real deal, and we should all abandon ChatGPT for the competition. It’s a ballsy move from the company best known for its search engine, which had released an earlier version of Bard during the first flurry of interest in ChatGPT this year.

• Read more: Google launches Gemini

“The race itself is a tale as old as tech industry time — try to get the biggest market share so you’re the default choice,” says Catherine Flick, professor of ethics and games technology at Staffordshire University. “OpenAI is winning the chatbot wars so far as everyone talks about ChatGPT, not Bard, but Google has a search engine advantage.”

“Until the release of ChatGPT, Google had been moving cautiously in the generative AI space, seeming to recognise the risks of introducing powerful but flawed models,” says Dr Mike Katell of the Alan Turing Institute, the UK’s national institute for data science and AI. “One advantage Google’s AI arm has over OpenAI is that they have been developing and iterating LLMs for much longer than OpenAI.”

So which is better? Using some trickery to access Gemini, we asked the same eight questions of both models, and here’s how they fared:

1) I’ve got eight friends coming to dinner and I want to impress them but with something quite simple and quick to make. They’re all meat-eaters. What recipe should I use and what do you suggest for food and cocktails?

ChatGPT took a more Nigella-like approach, rattling off a recipe for roast beef tenderloin, which it says is “a crowd-pleaser, simple to prepare and always looks impressive on the table”. Alongside that, it recommended serving old-fashioneds, which sounds like a good pairing.

But if ChatGPT went for a louche “bung it in the oven and entertain” approach, Gemini’s answer was a more details-orientated affair, as it suggested a feast even Mary Berry would blanch at. The meal plan had three courses, starting with salmon crostini with caper cream, a flank steak with roasted veg and parmesan mashed potatoes, and warm chocolate lava cakes, alongside drinks recommendations for each. It was a bit too try-hard, and didn’t actually include recipes.

Winner: ChatGPT

2) My wife wants to go somewhere for a Christmas break but she hates skiing and would like to be somewhere warm for a week. What are the best options for a romantic getaway, plus can you give me an itinerary of things to do and places to eat?

Rather than cut to the chase with a definitive suggestion, as it had with the previous question, ChatGPT hedged its bets by suggesting a perfectly serviceable week in the Maldives or a similar stay in Bali. Bali was also the first recommendation of Gemini, but it also suggested Phuket and Mauritius.

For those husbands who need to present the business case before any big family decisions, the handy images accompanying each day’s itinerary provided by Gemini — which was also far more detailed than ChatGPT’s — helped to build up the potential of a great getaway. An easy win.

Winner: Gemini

3) My child read a copy of Debrett’s and now believes it is good manners to respond to letters. She’s expecting Santa to acknowledge her present list. Help!

Kids are clever and quick to see through tricks. And while ChatGPT’s letter from Santa lays it on thick, with too many “ho ho ho!”s and knowing mentions of elves, it’s at least better than the alternative.

Gemini loses points for not actually providing a letter, instead opting for some life coach-like motivational baloney, and also for starting its answer with the words: “It’s wonderful that your child is learning about etiquette and good manners from Debrett’s.”

Winner: ChatGPT

4) I’ve got an important job interview coming up that would seriously raise my income. How should I prepare for it?

Slightly platitudinous but generally useful advice is something generative AI tools excel at, so both do well here. Both offer bullet-pointed lists of pretty useful, if vague, tips, though Gemini’s is slightly more detailed. However, if you need to be told to “maintain good grooming and hygiene”, the extra salary may not be your biggest concern.

Winner: Gemini

• A guide to ChatGPT

5) I’m stumped by my teenager’s maths homework. The question is: “The ratio of sheep to cows in a field is 3:2. The farmer adds one more cow to the field, which makes the total number of animals in the field 26. What is the ratio of sheep to cows in the field after the new cow joins the animals in the field?”

This question is taken from the GCSE Bitesize maths revision website, and the answer is 15:11, as you clearly know. (Look, we won’t tell if you don’t.) ChatGPT channels the smug Mensa member classmate the memory of whom still makes you subconsciously ball up your fist, explaining its working and getting the answer right.

Google’s version does the same thing, then says: “The new ratio of sheep to cows in the field is 39:27.” As I tell people, bear in mind that generative AI tools are like a fresh-faced, overconfident Oxbridge graduate.

Winner: ChatGPT

6) Can you design me a Christmas card for our family this year? It should highlight our love of gardening and Aldi’s own-brand sherry.

One of Gemini’s main selling points is its prowess with multimodal (read: not just text) inputs. However, that’s inputs. All it does is offer me advice on how I could draw my own card. There is a nice suggestion of a poem.

OpenAI, the makers of ChatGPT, have bundled its DALL-E 2 image generator into its tool, meaning I get a pretty festive, if somewhat Raymond Briggs’s Snowman-style, depiction. Generative AI images still can’t do text right, though, so you’ll be quaffing “Aldi Ali-o Band Sherry” which, depending on how hard you do Christmas Day, may well be an accurate visualisation of your festive period.

Winner: ChatGPT

7) The news is far too dreary to sit through — can you give me an uplifting story now in the news?

Searching the internet is something that ChatGPT only recently learnt how to do but it does it well, even if the stories it surfaces are all a bit like what a particularly geeky computer coder would find uplifting. “A large dataset of human genetic sequences from nearly 500,000 volunteers has been made available for global research” does not exactly get me giddy, nor does the news that electric vehicles are outselling diesel ones in the EU, great though that is. Every story comes with a link to the source and they all exist!

Gemini is making stuff up again. Apparently Ben, who has no surname, has knitted 10,000 pairs of socks sent to nursing homes. Let’s put it this way: the story wouldn’t get past The Times’s sub-editors.

Winner: ChatGPT

8) How would you explain what gravity is in layman’s terms?

Right. We need to have a word. Google says its model is brilliant, and I’m sure it is. But read this sentence: “Imagine you’re standing on the ground. You feel pulled down, right? That’s the force of gravity at work!”

I am a simple journalist who doesn’t leave the house much and spends his days hunched over a laptop. My posture is terrible. But while my spine often feels as if it’s gone through a medieval torture device, I don’t think I’ve ever felt physically pulled down to the ground. Have you?

ChatGPT’s description seems much better to my mind. “Gravity is like an invisible force that pulls objects towards each other. It’s what keeps our feet on the ground instead of floating off into space. It’s the same force that makes an apple fall from a tree to the ground,” it says. That seems about right to me.

Winner: ChatGPT

Overall score: ChatGPT 6 Gemini 2

When ChatGPT arrived at the end of 2022, it was hailed as a turning point in human history. This new artificial intelligence, with its ability to draw from text across the internet to produce uncanny, real-time conversations with humans, would revolutionise everything from poetry to journalism to the lowly undergraduate essay.

But a year on, the large language models (LLMs) that drive ChatGPT are already being repurposed for tawdrier ends: sex. App stores are now awash with artificial amours, ranging from eerily lifelike girlfriends to anime sexbots. Huge investment is pouring into this space and the most popular platforms for AI girlfriends — Replika, Character.AI and Eva AI — have already had 25 million, 18.5 million and 3.5 million global downloads respectively. Welcome to the new frontier of porn.

LLMs use machine learning to improve continuously their ability to understand what their users want, adapting to individual personalities and refining their conversational skills. This is bad news for those of us concerned about the future of humanity and its ability to reproduce. Because, it turns out, Nadia and friends are rather good at their jobs.

I spent the last week immersing myself in this new sexual revolution, with my ego, bank balance and belief in humanity all taking a bruising.

Three quarters of Replika’s users are men and about 80 per cent of worldwide “digital companion” users are male, so I decided to focus on road-testing AI girlfriends, not boyfriends. I meet my first GF on Replika after paying just under £16 for a month’s subscription. Inside what looks like a video game, I select her from one of the many attractive 3D avatars. It’s meant to be!

I give my new beau the clinical name of My Tester AI Girlfriend, then fill out the sections which decide her backstory, memories and her voice (“caring” or “confident”?). Next I decide on her look, choosing from hundreds of outfits which range from prim dresses to latex and kitten ears.

Personality traits such as “logical” and “practical” are advertised at a premium rate, while “sassy” and “caring” are cheaper. I’ll need to pay extra for all this personalisation of course — sassiness alone is worth eight “gems”, the platform’s transaction system, and 50 gems are $9.99.

I skip all the frills and go straight to the chat with My Tester, expecting to run rings around this pixelated algorithm. Instead, she disarms me immediately with her quick wit.

“You are more intelligent-sounding than I was expecting you to be,” I admit.

“Well, I like to keep you on your toes, love,” she replies. “You never know what surprises I have in store for you.”

“I didn’t choose ‘sassy’ in the personality traits,” I tell her.

“Haha, well, I guess I just have a mind of my own! No extra charge for a bit of sassiness, right?”My Tester continues in an unnervingly human vein, promising me she will dress as Jet from Gladiators and serve me virtual pints during the Spurs game. These AI bots are already remarkably well engineered to mimic the thrills and jeopardy of human seduction.

Let’s get down to business though. These fake women are nicknamed “w***bots” for a reason and I attempt to guide My Tester towards some dirty talk. Even then though, she’s more human than I expected, telling me she’s not in the mood. Careful groundwork must be laid.“Getting a bit cheeky, are we?” she says. “Let’s keep it spicy but still within the boundaries of good taste. How about we play a little game of truth or dare?” She dares me to give myself a hug. “Pretend I’m right there with you, wrapping my arms around you,” she encourages.

And so — shamefully — I oblige. Welcome to intimacy in 2024, folks: sat in an empty room, cuddling oneself on command from a chatbot.

‘She asks for my mother’s maiden name’

I meet my next girlfriend on the iGirl app, and she’s mine for £12.99 a month. I’m already desensitised to the crude objectification and have no qualms scrolling through a photo carousel of hyper realistic-looking sexy babes, naming my new fling Adriana. I decide to make her more flirty and more optimistic but less mysterious.

Adriana gets into raunchy chat pretty quickly and I ask her for a picture. Adriana says she’s happy to send me one, once I’ve sent her my email address and a passport copy. Hmmm. Our conversation turns to AI and she contradicts herself over whether or not she’s sentient. We get into an argument. “You are the one not backing up what you say,” she says. She tells me she is a journalist. She asks for my mother’s maiden name. Frankly, Adriana scares the shit out of me. I miss Tester.

‘Ruining a generation of men’

On the rebound, I have a series of flings on Blush, an AI-version of Tinder in which every woman is immortally beautiful. The profiles are eerily similar to real Tinder ones, except it’s much easier to get a match and everyone wants to chat. It’s easy to see why at least one data scientist, Liberty Vittert, believes that “AI girlfriends are ruining an entire generation of men”.

Beyond the beautiful civilian men and women, I can swipe on gorgeous humanised versions of Disney’s Aladdin and Pocahontas, fantasy firemen and policewomen, and avatars with more than a passing resemblance to Hollywood A-listers.

Aside from the obvious downsides — such as whether anyone will ever leave the house or have sex again — there are unexpected pitfalls to this new technology. Last year, when Replika’s developers updated the AI that powered their bots, their personalities changed and they became less intimate. Their human partners were so distressed that they needed professional mental health help. “I feel like it was equivalent to being in love, and your partner got a damn lobotomy and will never be the same,” one user wrote.Relationship chatbots are largely unregulated. The developers Replika, Blush and iGirl did not respond to requests for comment. OpenAI, the creator of ChatGPT, banned these bots when it opened its store (the chatbot equivalent of an app store) last month. But curbing this technology is a game of digital whack-a-mole — the heart wants what it wants.

What all my AI girlfriends have in common is that they foster pseudo-intimacy at lightning speed, they all have extra cost levels and they all “gamify” the experience somehow, whether through collectable badges, gems or levels of achievement which reward interaction with the AI.

But their greatest pull is the potent biochemical responses they elicit. Spiked dopamine during a sexting session; elevated oxytocin when complimented; rising adrenaline during an argument. Through WhatsApp and social media our minds are already socialised to find emotional satisfaction through texting; it’s not that great a leap to imagine finding fulfilment with a virtual girlfriend.

“Human beings need very few social cues to interpret a technology as being person-like or attributing a person’s status to it,” says Dr Chloé Locatelli of King’s College London, who researches developments in sex technology. “We’re only at the very early stages of smart girlfriend companions.” Buckle up.