Large artificial intelligence models will only get “crazier and crazier” unless more is done to control what information they are trained on, according to the founder of one of the UK’s leading AI start-ups.
Emad Mostaque, CEO of Stability AI, argues continuing to train large language models like OpenAI’s GPT4 and Google’s LaMDA on what is effectively the entire internet, is making them too unpredictable and potentially dangerous.
“The labs themselves say this could pose an existential threat to humanity,” said Mr Mostaque.
Today Sir Antony Seldon, headteacher of Epsom College, told Sky News’s Sophy Ridge on Sunday that AI could be could be “invidious and dangerous”.
Image: ‘Painting of Edinburgh Castle’ generated by artificial intelligence tool Stable Diffusion, whose founder warns not all internet users will be able to distinguish between real and AI images. Pic: Stable Diffusion
Image: An image of ‘print of fruits in green and orange’ generated by artificial intelligence tool Stable Diffusion, which converts text to images. Pic: Stable Diffusion
“When the people making [the models] say that, we should probably have an open discussion about that,” added Mr Mostaque.
But AI developers like Stability AI may have no choice in having such a discussion. Much of the data used to train their powerful text-to-image AI products was also “scraped” from the internet.
More on Artificial Intelligence
Related Topics:
That includes millions of copyright images that led to legal action against the company – as well as big questions about who ultimately “owns” the products that image- or text-generating AI systems create.
His firm collaborated on the development of Stable Diffusion, one of the leading text-to-image AIs. Stability AI has just launched a new model called Deep Floyd that it claims is the most advanced image-generating AI yet.
Advertisement
Image: A ‘photo of a fuzzy cute owl drinking very dark beer’ created by AI. Pic: DeepFloyd
Image: A photo-realistic style image of a ‘playful furry fox working as a pilot’ created by artificial intelligence. Pic: DeepFloyd
A necessary step in making the AI safe, explained Daria Bakshandaeva, senior researcher at Stability AI, was to remove illegal, violent and pornographic images from the training data.
If the AI sees harmful or explicit images during its training, it could recreate them in its output. To avoid this, the developers remove these images from the training data, so the AI cannot “imagine” how they would look.
But it still took two billion images from online sources to train it. Stability AI says it is actively working on new datasets to train AI models that respect people’s rights to their data.
Stability AI is being sued in the US by photo agency Getty Images for using 12 million of its images as part of the dataset used to train its model. Stability AI has responded that rules around “fair use” of the images means no copyright has been infringed.
But the concern isn’t just about copyright. Increasing amounts of data available on the web whether it’s pictures, text or computer code is being generated by AI.
“If you look at coding, 50% of all the code generated now is AI generated, which is an amazing shift in just over one year or 18 months,” said Mr Mostaque.
And text-generating AIs are creating increasing amounts of online content, even news reports.
Image: Image of ‘England wins men’s football world cup in 2026’ generated by artificial intelligence tool Stable Diffusion, which converts text to image, shows that the tool does not always get it spot on. Pic: Stable Diffusion
Please use Chrome browser for a more accessible video player
8:58
Sir Anthony Seldon highlights benefits and risks of AI
US company News Guard, which verifies online content, recently found 49 almost entirely AI generated “fake news” websites online being used to drive clicks to advertising content.
“We remain really concerned about an average internet users’ ability to find information and know that it is accurate information,” said Matt Skibinski, managing director at NewsGuard.
AIs risk polluting the web with content that’s deliberately misleading and harmful or just rubbish. It’s not that people haven’t been doing that for years, it’s just that now AI’s might end up being trained on data scraped from the web that other AIs have created.
All the more reason to think hard now about what data we use to train even more powerful AIs.
“Don’t feed them junk food,” said Mr Mostaque. “We can have better free range organic models right now. Otherwise, they’ll become crazier and crazier.”
A good place to start, he argues, is making AIs that are trained on data, whether it’s text or images or medical data, that is more specific to the users it’s being made for. Right now, most AIs are designed and trained in California.
“I think we need our own datasets or our own models to reflect the diversity of humanity,” said Mr Mostaque.
“I think that will be safer as well. I think they’ll be more aligned with human values than just having a very limited data set and a very limited set of experiences that are only available to the richest people in the world.”
The family of a father shot dead in a suspected case of mistaken identity in north London have said he “deserves justice” as they appealed for information.
Mahad Abdi Mohamed, 27, died from a gunshot wound to the head in hospital after he was hit with bullets fired from a stolen Mitsubishi Outlander, which was later found burnt out.
Detectives believe those responsible for his murder had set out to hurt someone else in a “pre-meditated and targeted attack” in Waverley Road, Tottenham, at 8.45pm on Thursday 20 March.
Mr Abdi Mohamed’s younger sister, Amal Abdi Mohamed, 23, said he was a “loving father” to his five-year-old son, who “looked up to him like a superhero”, and was planning to get married in the summer.
Image: Mahad Abdi Mohamed with his sister, Amal Abdi Mohamed. Pic: Met Police
“He was taken away from us through gun violence,” she said.
“A bullet didn’t just take his life, it tore through our family, through our heart, and it’s truly shocking, it’s devastating, and it’s so senseless, because this type of violence should never be normal.
“It should never be something a family ever has to expect, prepare for, or live with.”
More on Crime
Related Topics:
Mr Abdi’s 26-year-old friend, with whom he had been breaking his Ramadan fast, was also shot in the leg and was treated in hospital for a wound police said was not life-changing.
The Metropolitan Police arrested four men on suspicion of murder, who have been released on bail pending further investigations.
Detectives are appealing for witnesses who saw a silver Mitsubishi Outlander in the area, which was found burnt out in Runcorn Close, the following morning.
Image: A Mitsubishi was found burnt out the following day. Pic: Met Police
“This tragic event and Mahad’s death, has had a profound impact on the community and all those who loved him. Someone out there knows what happened. And that person, or people, must come forward,” said Detective Chief Inspector Rebecca Woodsford.
“Regardless of how small you think your information is, please share it with us. It could be the missing link we need to secure justice for Mahad and his family.”
Many of Mr Abdi Mohamed’s family members were in tears as they visited the scene of his murder as part of the appeal for information.
“My sweet Mahad was the kind of person who could light up a room without even trying,” said his sister.
“His laugh was so loud, and it still echoes in our memories.”
Ms Abdi Mohamed said her brother “was funny, he was honest, and overall he was just a good man” but “wasn’t perfect”.
She said he had “made mistakes but turned his life around” working at Waterloo Station, and part-time at the Tottenham Hotspur Stadium and Royal Ascot as a security guard.
Image: Mr Abdi Mohamed with his mother, Zahra Ali Seef. Pic: Met Police
“How do you look at a child who adored him day and night, and tell them that he’s gone and you don’t have the answers why? That boy will have to grow up with no dad,” she said.
“If you think you may know anything or have seen anything – you may think it doesn’t matter, but it might be the key to giving us an answer, and it might be the thing that finally lets our family take a breath.
“To stay silent is to be complicit.
“To stay silent is to let a grieving mother suffer in confusion. To stay silent is to let a little boy grow up not knowing what happened to his father.
“If you know something and you haven’t come forward, please think about that. Think about a family that cannot begin to heal because the truth is still hiding in the shadows. My brother deserves better. He deserves justice.”
Sir Alan Bates has called for those responsible for the wrongful convictions of sub postmasters in the Capture IT scandal to be “brought to account”.
It comes after Sky News unearthed a report showing Post Office lawyers knew of faults in the software nearly three decades ago.
The documents, found in a garage by a retired computer expert, describe the Capture system as “an accident waiting to happen”.
Please use Chrome browser for a more accessible video player
11:28
Post Office: The lost ‘Capture’ files
Sir Alan said the Sky News investigation showed “yet another failure of government oversight; another failure of the Post Office board to ensure [the] Post Office recruited senior people competent of bringing in IT systems” and management that was “out of touch with what was going on within its organisation”.
The unearthed Capture report was commissioned by the defence team for sub postmistress Patricia Owen and served on the Post Office in 1998 at her trial.
It described the software as “quite capable of producing absurd gibberish” and concluded “reasonable doubt” existed as to “whether any criminal offence” had taken place.
Ms Owen was found guilty of stealing from her branch and given a suspended prison sentence.
She died in 2003 and her family had always believed the computer expert, who was due to give evidence on the report, “never turned up”.
Image: Patricia Owen (right) was convicted in 1998 of stealing from her post office branch. She died in 2003
Adrian Montagu reached out after seeing a Sky News report earlier this year and said he was actually stood down by the defending barrister with “no reason given”.
The barrister said he had no recollection of the case.
Victims and their lawyers hope the newly found “damning” expert report, which may never have been seen by a jury, could help overturn Capture convictions.
Please use Chrome browser for a more accessible video player
1:49
What is the Capture scandal?
‘These people have to be brought to account’
Sir Alan, the leading campaigner for victims of the Horizon Post Office scandal, said while “no programme is bug free, why [was the] Post Office allowed to transfer the financial risk from these bugs on to a third party ie the sub postmaster, and why did its lawyers continue with prosecutions seemingly knowing of these system bugs?”
He continued: “Whether it was incompetence or corporate malice, these people have to be brought to account for their actions, be it for Capture or Horizon.”
More than 100 victims have come forward
More than 100 victims, including those who were not convicted but who were affected by the faulty software, have so far come forward.
Capture was used in 2,500 branches between 1992 and 1999, just before Horizon was introduced – which saw hundreds wrongfully convicted.
The Criminal Cases Review Commission (CCRC), the body responsible for investigating potential miscarriages of justice, is currently looking at a number of Capture convictions.
A CCRC spokesperson told Sky News: “We have received applications regarding 29 convictions which pre-date Horizon. 25 of these applications are being actively investigated by case review managers, and two more recent applications are in the preparatory stage and will be assigned to case review managers before the end of June.
“We have issued notices under s.17 of the Criminal Appeal Act 1995 to Post Office Ltd requiring them to produce all material relating to the applications received.
“To date, POL have provided some material in relation to 17 of the cases and confirmed that they hold no material in relation to another 5. The CCRC is awaiting a response from POL in relation to 6 cases.”
A spokesperson for the Department for Business and Trade said: “Postmasters negatively affected by Capture endured immeasurable suffering. We continue to listen to those who have been sharing their stories on the Capture system, and have taken their thoughts on board when designing the Capture Redress Scheme.”
Baroness Harman was solicitor general when Sir Tony Blair decided to take Britain to war in Iraq alongside the United States in 2003.
She said the decisions made by Sir Tony would be “burning bright” in Sir Keir‘s mind.
Image: Tony Blair’s decision to go to war in Iraq will be ‘burning bright’ in the PM’s mind, Baroness Harman said. Pic: PA
“He’s part of the political generation of the Labour Party that grew up, which was shaped by its opposition to what Tony Blair was doing in relation to Iraq,” Baroness Harman said.
“So it would be a massive change for him.”
More on Donald Trump
Related Topics:
Asked if the UK could end up giving permission for US aircraft to use British military bases on Cyprus and Diego Garcia, but not go any further than that, Baroness Harman said: “Exactly”.
Image: Soldiers guard the security gate of RAF Akrotiri, a British military base in Cyprus. File pic: Reuters
Sky News reported on Thursday that Attorney General Richard Hermer has raised questions over whether Israel’s actions in Iran are lawful, potentially limiting what support he believes the UK could offer the US.
Baroness Harman said that for Sir Keir, the “rules-based international order is the most important thing”.
“If the attorney general says that the government can’t do something because it’s illegal, it can’t do it. So he’s in a very crucial position,” she added.
Image: Harriet Harman (R) with Beth Rigby
But Baroness Harman said it would be difficult for Sir Keir to say “thank you for the trade deal” to Mr Trump and then deny the president use of the airbases.