Connect with us

News desk

AI generator Midjourney pauses service over deepfake ‘abuse’

Published

on

Hand-tracking tech firm Leap Motion co-founder David Holz went on to establish the Midjourney research lab causing a stir with image-generating software used to make deepfakes that have gone viral
Share this:

Research laboratory Midjourney has paused free trials of its image-generation software after users cranked out realistic deepfakes including of former US president Donald Trump getting arrested and Pope Francis in a puffer jacket.

Midjourney responded to a request Thursday for a trial with a message saying it could not be provided and to try again another day.

The imagery created using the artificial intelligence platform, particularly those of Trump and the pontiff which went viral, have put a spotlight on the San Francisco-based lab.

“Due to a combination of extraordinary demand and trial abuse we are temporarily disabling free trials until we have our next improvements to the system deployed,” Midjourney founder David Holz said in a post this week on the company’s Discord channel.

The service generates realistic looking images based on written prompts made by users.

It launched in test mode in mid-2022, with the independent lab consistently upgrading the software.

Users have praised a freshly released version of Midjourney for improved realism in produced images.

Along with putting the brakes on new free trials, Midjourney banned certain words, such as “arrested,” from being used to prompt image creation.

On Thursday Midjourney denied a request by AFP to generate an image of the former president being arrested in front of Trump Tower in New York.

“The word ‘arrested’ is banned,” a message from Midjourney stated.

“Circumventing this filter to violate our rules may result in your access being revoked.”

Billionaire mogul and Twitter owner Elon Musk and a range of experts called on Wednesday for a pause in the development of powerful artificial intelligence systems to allow time to make sure they are safe.

An open letter, signed by more than 1,000 people so far including Musk and Apple co-founder Steve Wozniak, was prompted by the release of artificial intelligence platform GPT-4 from Microsoft-backed firm OpenAI.

Canadian AI pioneer Yoshua Bengio, who signed the letter, warned during a virtual press conference in Montreal that “society is not ready” for this powerful tool, and its possible misuses.

“Let’s slow down. Let’s make sure that we develop better guardrails,” he said.

Share this:

News desk

California looks to Europe to rein in AI

Published

on

By

Legislators in the California state capitol are working on a flurry of laws aiming to crack down on abusive uses of artificial intelligence on the home turf of some of the world's powerful tech titans
Share this:

California, home to Silicon Valley, is eager to rein in the deployment of artificial intelligence and is looking to Europe’s tough-on-big-tech approach for inspiration.

The richest state in the United States by GDP, California is a hotbed of no-holds-barred tech innovation, but lawmakers in state capital Sacramento want to give the industry laws and guardrails it has largely been spared in the internet age.

Brussels has enacted a barrage of laws on US-dominated tech and sprinted to pass the AI Act after OpenAI’s Microsoft-backed ChatGPT arrived on the scene in late 2022, unleashing a global AI race.

“What we’re trying to do is actually learn from the Europeans, but also work with the Europeans, and figure out how to put regulations in place on AI,” said David Harris, senior policy advisor at the California Initiative for Technology and Democracy.

As they have in the past with EU laws on private data, lawmakers in California are looking to recent European legislation on AI, especially given the little hope of equivalent national legislation out of Washington.

There are at least 30 different bills proposed by California state legislators that relate to various aspects of AI, according to Harris, who said he has advised officials here and in Europe on such laws.

Proposed laws in California range from requiring AI makers to reveal what was used to train models to banning election ads containing any computer generated features.

“One of the aspects I think is really important is the question of how we deal with deepfakes or fake text created to look like a human being is sending you messages,” Harris told AFP.

State assembly member Gail Pellerin is backing a bill she says would essentially ban the spreading of deceptive digital content created with generative AI in the months leading up to and the weeks following an election.

“Bad actors who are utilizing this are really hoping to create chaos in an election,” Pellerin said.

– Law-breaking ‘bad guys’ –

Industry association NetChoice is dead set against importing aspects of European legislation on AI, or any other EU tech regulation.

“They are taking, essentially, a European approach on artificial intelligence – which is that we must ban the technology,” said Carl Szabo, the general counsel of the association, which advocates for light touch regulation of tech.

“Outlawing AI won’t stop (anything). It’s bad because bad guys don’t follow the law,” Szabo argued.

“That’s what makes them bad guys.”

US computer software giant Adobe, like most tech giants, worked with Europe on the AI Act, according to Adobe General Counsel and Chief Trust Officer Dana Rao.

At the heart of the EU AI Act is a risk-based approach, with AI practices deemed more risky getting more scrutiny.

“We feel good about where the AI Act ended up” with its high-risk, low-risk approach, said Rao.

Already, Adobe engineers carry out “impact assessments” to rate risk before making AI products available, according to Rao.

“You want to think about nuclear safety, about cybersecurity, about when AI is making substantial decisions over human rights,” Rao said.

– ‘Watching California’ –

In California, Rao said he expected the problem of deepfakes to be the first to fall under the authority of a new law.

Assembly Bill 602 would criminalize non-consensual deepfake pornography while Assembly Bill 730 bans the use of AI deepfakes during election campaign season.

To fight this, Adobe joined other companies to create “content credentials” that Rao equated to a “nutrition label” for digital content.

Assemblywoman Pellerin expects AI laws adopted in California to be replicated in other states.

“People are watching California,” Pellerin said, with a slew of US states also working on their own AI deepfake bills.

“We’re all in this together; we have to stay ahead of the folks that are trying to wreak havoc in an election,” she said.

Share this:
Continue Reading

News desk

Big tech told to identify AI deepfakes ahead of EU vote

Published

on

By

EU parliamentary elections are taking place in the bloc's 27 member states on June 6-9
Share this:

The EU called on Facebook, TikTok and other tech titans on Tuesday to crack down on deepfakes and other AI-generated content by using clear labels ahead of Europe-wide polls in June.

The recommendation is part of a raft of guidelines published under a landmark content law by the European Commission for digital giants to tackle risks to elections including disinformation.

The EU executive has unleashed a string of measures to clamp down on big tech, especially regarding content moderation.

Its biggest tool is the Digital Services Act (DSA) under which the bloc has designated 22 digital platforms as “very large” including Instagram, Snapchat, YouTube and X.

There has been feverish excitement over artificial intelligence since OpenAI’s ChatGPT arrived on the scene in late 2022, but the EU’s concerns over the technology’s harms have grown in parallel.

Brussels especially fears the impact of Russian “manipulation” and “disinformation” on elections taking place in the bloc’s 27 member states on June 6-9.

In the new guidelines, the commission said the largest platforms “should assess and mitigate specific risks linked to AI, for example by clearly labelling content generated by AI (such as deepfakes)”.

The commission recommends that big platforms promote official information on elections and “reduce the monetisation and virality of content that threatens the integrity of electoral processes” to diminish any risks.

“With today’s guidelines we are making full use of all the tools offered by the DSA to ensure platforms comply with their obligations and are not misused to manipulate our elections, while safeguarding freedom of expression,” said the EU’s top tech enforcer, Thierry Breton.

While the guidelines are not legally binding, platforms must explain what other “equally effective” measures they are taking to limit the risks if they do not adhere to them.

The EU can ask for more information and if regulators do not believe there is full compliance, they can hit the firms with probes that could lead to hefty fines.

– ‘Trusted’ information –

Under the new guidelines, the commission also said political advertising “should be clearly labelled as such” before a tougher law on the issue comes into force in 2025.

It also urges platforms to put in place mechanisms “to reduce the impact of incidents that could have a significant effect on the election outcome or turnout”.

The EU will conduct “stress-tests” with relevant platforms in late April, it said.

X has already been under investigation since December over content moderation.

And the commission on March 14 pressed Facebook, Instagram, TikTok and four other platforms to provide more information on how they are countering AI risks to polls.

In the past few weeks, several of the companies including Meta have outlined their plans.

TikTok on Tuesday announced more of the measures it was taking including push notifications from April that will direct users to find more “trusted and authoritative” information about the June vote.

TikTok has around 142 million monthly active users in the EU — and is increasingly used as a source of political information among young people.

Share this:
Continue Reading

News desk

Ousted WeWork co-founder bids to buy company: reports

Published

on

By

The pandemic exacerbated WeWork's woes as people avoided offices for fear of Covid-19
Share this:

Ousted WeWork co-founder Adam Neumann recently bid more than $500 million to buy back the struggling office-sharing group, according to media reports Monday.

Neumann recently submitted the offer, but it was not clear how he would finance such a deal, the Wall Street Journal and CNBC reported, citing sources close to the matter.

Neumann is seeking to buy the company out of bankruptcy, according to a letter to WeWork seen by AFP last month.

WeWork went into bankruptcy in November with its major creditors set to take control of the company.

At its height, WeWork was the biggest private renter of office space in Manhattan, with co-working spaces in cities across the globe.

But investors became concerned not only about WeWork’s business model and unbridled growth, but also about Neumann’s reliability as a boss.

A charismatic figure, Neumann was known for his sometimes abrupt decisions.

In September 2019, the Board of Directors dismissed him, shortly after the company’s stock market debut was postponed. 

Neumann was forced out of the company with a $1.7 billion exit package, while the company’s value was slashed to $8 billion.

WeWork was then disrupted by the Covid-19 pandemic, which emptied offices as workers went remote, and the company never fully recovered.

Share this:
Continue Reading

Featured