Companies developing AI models today are in a Darwinian competition for market share and profits. Be it Google or Microsoft, the mantra seems to be: ethics be damned. Both the big tech companies with billions of dollars at stake in the game have trimmed their responsible and ethical AI teams in 2023, yet no signs of pause or shuffle is visible among the other AI teams. 

For Sasha Luccioni, an AI researcher and climate lead at Hugging Face, the distinction between “responsible AI” and just plain AI is bothersome. “It makes no sense. It’s like you imagine there being safe cars and cars,” she quipped. Reflecting on the layoffs, she surmised there were budget cuts and recession, but lack of agreement between the teams was probably the biggest reason they were let go.

“When you have this distinction, there’s too much friction because the responsible AI team’s job is essentially to push back,” Luccioni mused. Recollecting the Google debacle two years ago, where the company’s star AI ethicist Timnit Gebru and her team faced the axe for sounding alarms on the dangers of large language models, Luccioni stated, “That is what they were hired to do, and yet when the push comes, responsible AI researchers are the ones that get shoved out because they’re in conflict with the broader profit model of the company.” 

Among the list of companies struggling to find a place for AI ethicists was Meta, the multinational conglomerate. The company run by Mark Zuckerberg dissolved its responsible AI division and transferred its (human) resources into different generative AI teams. Luccioni sees merit in the strategy and calls it a logical move. “You shouldn’t have an isolated responsible AI team; you should have an integrated responsible AI network,” she contended. 

Similarly, Hugging Face’s responsibility experts regularly meet people across the company who work on different projects, so there’s no tension between the responsible team and the rest of the company. “That means from the beginning of a project, we are thinking about responsibility and is not just something added as an appendix at the end,” Luccioni proudly noted.

Generative AI: Interesting, not Revolutionary

What frustrates Luccioni, she says, is the fact that AI has become so much about marketing and who makes the most noise with a launch. “The focus is so less on robustness, transparency or responsibility. Even if a model is launched, there’s always some fine print,” she pinpointed and quoted the recent example of Google launching Gemini through a demo mired in controversy due to edits. 

This was one of the company’s gambit gone wrong to gain an edge in generative AI. Last year, Google entered panic mode after ChatGPT became the most popular kid on the block and ended up losing a $100 billion valuation after Bard’s live demo spat out factually inaccurate answers.

At the mere mention of the technology, Luccioni’s scepticism is palpable. “Generative AI is interesting,” she acknowledged, “I just don’t see the commercial advantages yet. Generating an image is great and could be useful, but it won’t change the world. Same for ChatGPT or large language models, they’re useful for different cases, but it’s not like this groundbreaking, revolutionary technology that people say it is,” she added.

Technologically, Luccioni is a big fan of old-school approaches. “I spent some time working with the United Nations and realised to what extent all the big models that we develop in the lab are not as useful in the real world,” she noted.  

Breaking The Binary

Apart from researching at Hugging Face, the 33-year-old is also a Founding Member of Climate Change AI. She has good enough reasons not to be impressed by the human-mimicking technology which has become the belle of the ball in Silicon Valley.  

Luccioni is not much impressed by the tech regarding climate change applications. Some researchers trained a model on IPCC reports that can reply to questions about global warming and the greenhouse effects. “But honestly, you can do the same thing with just information retrieval; you don’t need generative AI.” Luccioni asserted. While acknowledging potential use cases, she dismissed it as a game-changer. “Especially given the amount of resources and the planetary impact of these models I don’t think it’s worth using them given their costs essentially,” she added with conviction. 

Luccioni suggested that impactful actions can be taken, like channelling AI expertise to support existing communities of climate change researchers. “But it’s not really like starting from scratch. They were doing the work before AI came along,” she said. “It’s the kind of work that’s hard to publish in conferences because it’s usually not novel or state of the art. But it can really help communities of scientists do their jobs more efficiently.”  

“When I started in AI, I felt that the work we were doing was really to make better and useful technologies that will help people,” Luccioni recalled. “I was just reading about Imagen 2, which is great research work, but why are we doing this? What problem is it solving?” she pondered. “I feel like the values of the field have shifted, and I find it difficult to get excited about this stuff anymore.” 

Seven years ago, Luccioni, an applied AI researcher in finance, pivoted to her true passion—climate and nature. She quit her job, took a massive pay cut and decided to use her AI background to help fight climate change. After a stint with computer scientist Yoshua Bengio, she sought a space between academia and the industry. 

“I got a couple of job offers from big tech companies, and then I chose Hugging Face because I believe in the mission and importance of keeping AI as accessible as possible to as many people,” she said. 

“That’s why I keep doing the work that I do because I want to understand it better, and it’s not a binary thing. It’s not like AI is good or bad for the planet. It’s more like there are ways in which AI can help. But to figure out whether it’s more good than bad, you need to understand its impact,” Luccioni said in conclusion.