[News] The AI Thread!

Once upon a time, all stories were written solely by humans. Now, researchers have found AI might help authors tell a tale.

A study suggests that ideas generated by the AI system ChatGPT can help boost the creativity of writers who lack inherent flair – albeit at the expense of variety.

Prof Oliver Hauser, a co-author of the research from the University of Exeter, said the results pose a social dilemma.

“It may be individually beneficial for you to use AI, but as a society if everyone used AI, we might all lose out on the diversity of unique ideas,” he said. “And, arguably, for creative endeavours we might sometimes need the ‘wild’ and ‘unusual’ ideas.”

The team asked 293 people to name 10 words that differed as much as possible from each other, allowing them to probe participants’ inherent creativity.

The researchers then randomly assigned participants one of three topics – an adventure in the jungle, on the open seas or on a different planet – and asked them to write an eight-sentence story appropriate for teenagers and young adults.

While a third of participants were offered no assistance, the others were split between those allowed to have one three-sentence starting idea pre-generated by ChatGPT, and those who could request five such ideas.

Overall, 82 of 100 participants took up the offer of a single AI-generated idea, while 93 of 98 participants offered access to five such ideas took at least one – and almost a quarter requested all five.

A further 600 participants, unaware of whether AI-generated ideas were used, read the resulting stories, and rated factors relating to novelty and usefulness – such as whether the story was publishable – on a nine-point scale.

The results, published in the journal Science Advances, reveal access to AI boosted these scores, with greater access associated with a larger effect: people with the option of five AI-generated ideas had an 8.1% increase, on average, in novelty ratings for their stories compared with people lacking the option of such help, while usefulness ratings rose by 9% on average.

“The effect sizes are not very large, but they were statistically significant,” said Dr Anil Doshi, a co-author of the study from University College London.

Stories written by people with the option of AI-generated ideas were also deemed more enjoyable, more likely to have plot twists and be better written.

However, it was writers with low inherent creativity that benefited most.

“We do not find that the most inherently creative people’s stories are being “supercharged” from AI ideas – this group of people is highly creative with and without the use of AI,” said Doshi.

The team also found participants with access to AI-generated ideas produced stories with greater similarity, something Doshi suggested is down to AI generating relatively predictable story ideas.

Hauser said such studies are important. “Evaluating the use of AI will be essential in making sure that we reap the benefits of this potentially transformative technology without falling prey to potential shortcomings,” he said.

The global rage for manga keeps on going with publishers unable to release translated editions as quickly as demand requires. This has fueled piracy as well as innovation in translation. NHK World's John LaDue went to find out what challenges lie ahead.

OpenAI is announcing its much-anticipated entry into the search market, SearchGPT, an AI-powered search engine with real-time access to information across the internet.

The search engine starts with a large textbox that asks the user “What are you looking for?” But rather than returning a plain list of links, SearchGPT tries to organize and make sense of them. In one example from OpenAI, the search engine summarizes its findings on music festivals and then presents short descriptions of the events followed by an attribution link.

In another example, it explains when to plant tomatoes before breaking down different varieties of the plant. After the results appear, you can ask follow-up questions or click the sidebar to open other relevant links. There’s also a feature called “visual answers,” but OpenAI didn’t get back to The Verge before publication on exactly how this works.

earchGPT is just a “prototype” for now. The service is powered by the GPT-4 family of models and will only be accessible to 10,000 test users at launch, OpenAI spokesperson Kayla Wood tells The Verge. Wood says that OpenAI is working with third-party partners and using direct content feeds to build its search results. The goal is to eventually integrate the search features directly into ChatGPT.

It’s the start of what could become a meaningful threat to Google, which has rushed to bake in AI features across its search engine, fearing that users will flock to competing products that offer the tools first. It also puts OpenAI in more direct competition with the startup Perplexity, which bills itself as an AI “answer” engine. Perplexity has recently come under criticism for an AI summaries feature that publishers claimed was directly ripping off their work.

A highly-praised AI video generation tool made by multi-billion dollar company Runway was secretly trained by scraping thousands of videos from popular YouTube creators and brands, as well as pirated films, according to a massive internal spreadsheet of training data obtained by 404 Media.

The model—initially codenamed Jupiter and released officially as Gen-3—drew widespread praise from the AI development community and technology outlets covering its launch when Runway released it in June. Last year, Runway raised $141 million from investors including Google and Nvidia, at a $1.5 billion valuation.

When Techcrunch asked Runway co-founder Anastasis Germanidis in June where the training data for Gen-3 came from, he would not offer specifics.

“We have an in-house research team that oversees all of our training and we use curated, internal datasets to train our models,” Germanidis told Techcrunch.

Big Tech's message to investors on back-to-back earnings calls this week was "Stop worrying about the billions we're spending on AI — everything's going to be just great."

The big picture: Over the past month, Wall Street began questioning the growing cost of buying high-end chips and building data centers, asking when these investments would start to show up in revenue and profit growth.

- Google, Microsoft and Meta, each in different ways, told Wall Street to relax.

By the numbers: These companies all continue to grow and remain jaw-droppingly profitable, providing an enviable financial foundation for their most ambitious dreams.

- Last week, Google parent Alphabet reported $24 billion net profit on $85 billion revenue.
- On Tuesday Microsoft reported $22 billion net profit on $65 billion revenue.
- On Wednesday Meta reported $13.5 billion net profit on $39 billion revenue.

Apple and Amazon — the two tech giants that aren't as directly in the "build the biggest and best foundation model" race but are both making their own significant AI investments — are no slouches, either.

- On Thursday Apple reported $21 billion net profit on $86 billion revenue.
- The same day, Amazon reported $13.5 billion net profit on $148 billion revenue.

Remember, all these numbers are for a single quarter — multiply by four for a very rough annual picture.

Follow the money: When an industry is geysering profits like this, shareholders can try to demand higher dividend payouts. Tech firms have traditionally shunned dividends as a mark of corporate old age, but some of the giants, like Apple, have warmed up to them.

- The other thing you can do with profit rivers and the cash reservoirs they build is plow them into the Next Big Thing — and for now, most investors seem happy to let CEOs take that approach.

Between the lines: Wall Street has already given all these companies massive valuations based on their current businesses and their long-term prospects.

- As with previous platform shifts in tech, though, the AI wave could crown new winners and drown current leaders.

CEOs hastened to reassure investors that the best is yet to come, and that their incumbent firms are well-positioned to control the future.

- AI will improve "almost every" existing Meta product and "make a whole lot of new ones possible," Meta CEO Mark Zuckerberg said on his call Wednesday. "So it's why there are all the jokes about how all the tech CEOs get on these earnings calls and just talk about AI the whole time. It's because it's actually super exciting."

As for doubts about the wisdom of plowing so many billions into "capex" (capital expenditures) to build AI capacity, Microsoft CEO Satya Nadella said the company is carefully tuning into the "demand signals" from customers.

- Those signals — like massive growth in use of GitHub Copilot, which is now a bigger business than GitHub itself was when Microsoft acquired it in 2018 — are flashing green, Nadella said.
- Executives also argued that if they overbuild data centers and buy more servers than the AI buildout ultimately needs, the companies will find other good uses for that infrastructure.

The bottom line: "When you go through a curve like this, the risk of underinvesting is dramatically greater than the risk of overinvesting for us here," Google CEO Sundar Pichai told analysts last week.

Meta is in discussions with Awkwafina, Judi Dench and other actors and influencers for the right to incorporate their voices into a digital assistant product called MetaAI, according to three people with knowledge of the talks, as the company pushes to build more products that feature artificial intelligence.

Apart from Ms. Dench and Awkwafina, Meta is in talks with the comedian Keegan-Michael Key and other celebrities, said the people, who spoke on the condition of anonymity because the discussions are private. They added that all of Hollywood’s top talent agencies were involved in negotiations with the tech giant.

The talks remain fluid, and it is unclear which actors and influencers, if any, may sign on to the project, the people said. If the parties come to an agreement, Meta could pay millions of dollars in fees to the actors.

A Meta spokesman declined to comment. The discussions were reported earlier by Bloomberg.

Meta, which owns Facebook, Instagram and WhatsApp, has invested heavily in artificial intelligence, which the biggest tech companies are racing to develop and lead. Meta has plowed billions into weaving the technology into its social networking apps and advertising business, including by creating artificially intelligent characters that could chat through text across its messaging apps.

On Wednesday, Mark Zuckerberg, Meta’s chief executive, increased how much his company would spend on A.I. and other expenses this year to at least $37 billion, up from $30 billion at the beginning of 2024. Mr. Zuckerberg said he would rather build too fast “rather than too late” to prevent his competitors from gaining an edge in the A.I. race.

One area of A.I. that is rapidly emerging are chatbots with voice abilities, which act as virtual assistants. In May, OpenAI, a leading A.I. company, unveiled a version of its ChatGPT chatbot that could receive and respond to voice commands, images and videos. It was part of a wider effort to combine conversational chatbots with voice assistants like the Google Assistant and Apple’s Siri.

OpenAI later suspended the release of its voice-related ChatGPT after the actress Scarlett Johansson, who had provided the voice for an A.I. system in Spike Jonze’s 2013 movie, “Her,” accused the company of using a voice “eerily similar to mine” despite her refusals to participate in the product.

Meta is angling to strike deals with celebrities in a way that avoids ticking off top talent. Under the terms of the proposed contract, Meta would record the voices of these celebrities for potential use in MetaAI, which users could interact with and ask questions across Facebook, Instagram, WhatsApp and Messenger, as well as Meta’s Ray-Ban augmented reality glasses, the people said. Any deal would be for a set period and could be renewed or terminated when the contract was up. Actors would not release the rights to their voices indefinitely.

Meta is trying to lock down the deals in time for its Connect technology conference in late September, when the company plans to debut new A.I.-focused products. At last year’s conference, Meta introduced digital chatbots that had the likenesses of Snoop Dogg, Tom Brady and MrBeast, but few people used the text-based characters. The company wound down the initiative last month.

Last year, the SAG-AFTRA union, which represents more than 150,000 television and movie actors, went on strike as negotiations over a new agreement with the Hollywood studios — including how to deal with the rise of A.I. — stalled. Actors eventually secured a three-year deal with a provision that says the studios cannot use A.I. tools to create digital replicas of performers without payment or approval.

Still, many union members remain dissatisfied with the provision. Other subsets of the entertainment industry, like editors, animators and voice actors, are also concerned that A.I. will put them out of work.

Actors with the union initiated a separate strike last week against video game companies that use actors’ images and voices in their games.

Replika’s basic pitch is pretty simple: what if you had an AI friend? The company offers avatars you can curate to your liking that basically pretend to be human, so they can be your friend, your therapist, or even your date. You can interact with these avatars through a familiar chatbot interface, as well as make video calls with them and even see them in virtual and augmented reality.

The idea for Replika came from a personal tragedy: almost a decade ago, a friend of Eugenia’s died, and she fed their email and text conversations into a rudimentary language model to resurrect that friend as a chatbot. Casey Newton wrote an excellent feature about this for The Verge back in 2015. Even back then, that story grappled with some of the big themes you’ll hear Eugenia and I talk about today: what does it mean to have a friend inside the computer?

That all happened before the boom in large language models, and Eugenia and I talked a lot about how that tech makes these companions possible and what the limits of current LLMs are. Eugenia says Replika’s goal is not to replace real-life humans. Instead, she’s trying to create an entirely new relationship category with the AI companion, a virtual being that will be there for you whenever you need it, for potentially whatever purposes you might need it for.

Right now, millions of people are using Replika for everything from casual chats to mental health, life coaching, and even romance. At one point last year, Replika removed the ability to exchange erotic messages with its AI bots, but the company quickly reinstated that function after some users reported the change led to mental health crises.

That’s a lot for a private company running an iPhone app, and Eugenia and I talked a lot about the consequences of these ideas. What does it mean for people to have an always-on, always-agreeable AI friend? What does it mean for young men, in particular, to have an AI avatar that will mostly do as it’s told and never leave them? Eugenia insists that AI friends are not just for men, and she pointed out that Replika is run by women in senior leadership roles. There’s an exchange here about the effects of violent video games that I think a lot of you will have thoughts about, and I’m eager to hear them.

Of course, it’s Decoder, so along with all of that, we talked about what it’s like to run a company like this and how products like this get built and maintained over time. It’s a ride.