Leaders like Elon Musk are calling for an AI pause. But experts say there's no secret to better, safer AI: Just add humans.

• Leaders like Elon Musk have called for a "pause" on AI development to better consider its effects on society.
• But industry insiders say that we already know the best way to make sure AI acts responsibly: Just add humans.
• The problem is that Big Tech giants have laid off huge chunks of their AI ethics teams as they push to cut costs.

In recent months, the discourse around artificial intelligence has hit something of a fever pitch.

Some of the loudest voices in tech, including Elon Musk and Steve Wozniak, warn that the rise of large language models like OpenAI's GPT-4 is leaving the door wide open for a new era where robots make all the decisions — if not the end of the human race altogether.

The conversation has now turned to the notion of responsible AI, as leaders like Elon Musk call for a temporary pause in the development of new AI models while we figure out how to make sure that these systems always act fairly and ethically in all situations.

The thing is that the secret to responsible AI is no secret at all. It's something that the industry has known for years. If you want to be completely sure that your AI is acting the way you want it to, you need a human in the loop.

"We need to have humans involved in the training, and even code ahead of time, how machines should act in times of a moral quandary," said Oded Netzer, a professor at Columbia Business School.

By making sure that the AI is coded for the desired behaviors from the jump, and checking in regularly to ensure that it's still acting as expected, it's largely accepted as standard practice that you can keep these systems from spewing misinformation, discriminating against certain users, or otherwise acting badly.

That's the good news. The bad news is that this sudden interest in responsible AI comes amid the recent onslaught of tech layoffs, as Big Tech companies like Microsoft, Google, and Meta cut their AI ethicists amid a broader drive for cost savings, as reported by the Washington Post.

It may have been the right move for their businesses in the short term, but those layoffs came at the worst possible time for the future of AI technology, the experts said.

"The last place you would want to let people go is in this domain," Netzer said. "I predict this may be a short-term low, and they will eventually rehire within these teams."

Human intervention is key for responsible AI

AI ethicists don't sit around a room discussing Plato and wondering what being human means. A lot of responsible AI deals with understanding the impact of AI in people's day-to-day lives. Think of AI ethicists as the responsible parent in the room.

Essentially, having an AI ethicist involved in the development process is intended to slow down the rush of getting a new product to market just because you can. Each company goes about things differently, but these ethicists generally check AI models for bias or inaccuracies and ask questions about the system's social or political impact. For example, they can ask product teams to hold back on release if they believe the AI can potentially spit out biased results or harm a group of people.

That's especially important in heavily regulated industries like finance or healthcare, where an unchecked algorithm could potentially reveal sensitive data if someone knows how to ask it the right (or wrong) question.

The ethicist doesn't even have to be someone who did spend their university years studying philosophy. Often, these are experts in their field, from lawyers (and other legal professionals) to people experienced in working with marginalized groups.

"I genuinely believe AI is the future, but the reality is, our digital lives should be built by geeks like me alone; it needs to be a combination of people coming from many different backgrounds who understand context and society," said Kriti Sharma, chief product officer of legal tech at Thomson Reuters.

She added that having an expert on tap while training AI gives the model better instructions and gives context to the data it takes in. With additional context, it helps cut down on bias and directs results.

All of which is to say, a way to create responsible AI already exists. No need for a six-month moratorium; just put humans in its way.