It’s Not 1995 Anymore: Craig Newmark on Why the US Needs to ‘Move Fast’ on AI

Craig Newmark is a rare tech billionaire who doesn’t see the value in status symbols like a Porsche or Rolex, a symptom of what he comfortably calls his “nerdly dysfunction.”

The 70-year-old programmer-turned-philanthropist stepped down as Craigslist CEO in 2000 to focus on the best ways to donate his fortune. Recently, he’s been working with organizations that are developing ethical AI guardrails, such a Common Sense Media and the Mozilla Responsible AI Challenge (with a $100 million donation to support veterans and military families sprinkled in).

As someone who was formerly at the center of another major technological change—the advent of the internet—he brings a unique perspective to the space. “Back then, we were more optimistic than we should have been,” Newmark said in a recent interview. “Today the spirit is different. It’s a more skeptical environment, and rightfully so.”

But rather than pausing AI development, as Elon Musk and other high-profile figures have suggested, Newmark advocates for the US to move quickly—and responsibly. AI is “on par with the invention of the printing press,” Newmark says. “The printing press changed the rules [and then] the internet changed the rules, so it’s both scary and a little bit exciting.”

‘We Need to Move Fast on AI’

Microsoft, Amazon, Dell, and others appearing on the 1999 NASDAQ stock exchange. (Credit: Kim Steele / Getty Images)

As Craigslist took off in the mid-’90s, Newmark found himself in the middle of an internet gold rush similar to what we see today with AI.

“The buzz of easy money for the right technology was in the air in general, and the subject of talk at industry parties and events,” he recalls. “There was this optimistic spirit that we would work together and make things better for everyone.”

Investors offered him “billions” to monetize Craigslist with ads and listing fees, but he turned them down. That decision was surprisingly easy. He felt the site was useful as-is, and, because of his “dysfunction,” saw no value in having more to spend on flashy items.

He has no regrets: “I’ve seen very frequently power makes you stupid, and more often than not, money makes you stupid.” The techno-optimism of the dot-com era also hasn’t proven to be foolproof. “Now that everyone has gotten on the net, we know there are a lot more people with good intentions, but just a few bad people have hurt us all in big ways,” he says.

Craigslist has struggled with content moderation, an area Newmark hopes AI can improve. He tried his hand at it for Craigslist, and the images still haunt him. He suspects those who do it for a living have a high likelihood of “ending up in therapy,” a main reason he funds anti-online harassment efforts. But an AI tool could remove the need for human review, provided the systems are cost-effective and have fewer false positives than today’s tools. OpenAI, for example, argues that its GPT-4 model speeds up the process with “less involvement from human moderators.”

OpenAI CEO Sam Altman (Credit: Tomohiro Ohsumi / Stringer / Getty Images)

Newmark may have turned down billions for Craigslist, but he says OpenAI CEO Sam Altman and those at the forefront of AI development will not have the same luxury. “We need to move fast in this area,” he says. “There are people and nation-states with bad intentions, and I do know people at the Defense Department, NSA, and FBI are real concerned about this.”

That includes China, which has a desire to “dominate the world,” he says. “They have lots of smart people and lots of money, but their economy is not in good shape. They’re factionalized, so different factions will do different things to compete with each other and get results faster. They have AI guys who just live in their offices and work 16 hours a day.”

President Biden issued an executive order this month to limit US investment in Chinese technologies, including AI, to curtail the competition. However, Newmark has little faith in Washington’s ability to regulate AI, as the EU has started to do, though he would like to be proven wrong. “The atmosphere is so toxic in Washington,” he says. “And I see politicians getting involved with AI as PR stunts, politicians who don’t understand this stuff.”

‘Don’t Lie to Me’ and 4 Other Must-Have AI Principles

Beginning a conversation with ChatGPT. (Credit: OpenAI)

If it was up to Newmark, every AI tool would follow a set of basic principles. Although he’s still educating himself on the intricacies of the technology, some ideas feel obvious based on the values instilled in him as a child. He also regularly meets with top government officials and experts on the subject, like his next-door neighbor, a Columbia University professor who likes to chat about neural networks over coffee.

1. No Disinformation—“The [number one thing] is, ‘Don’t lie to me.’ Thou shalt not bear false witness. The Bible tells us it’s a good guardrail, and in Sunday School that’s what Mr. and Mrs. Levin told me.”

2. Train on Quality Sources—“For example, The New York Times and Washington Post might not be perfect, but at least they try to get it right, whereas publications like Russia Today exist to lie to people. It’s not rocket science.”

3. Source Attribution—“Any AI generating results should credit the source. Because that’s what people should do for each other. That’s treating people like you want to be treated, and it’s fair.”

4. Opt-Out Options—“[AI tools] should make it easy to turn off their recommendation, or shorten it so you can see more alternatives.”

5. Transparency—“[AI tools] should explain to the best amount possible how they came to their conclusion.”

As noted in PCMag’s reviews of the three main AI chatbots—ChatGPT, Google Bard, and Bing Chat—none achieve all five. For example, ChatGPT does not attribute its information, which would have been helpful for the recipient of an off-the-rails response that deemed having a child “illegal.”

Newmark also calls for more neurology research, which can reveal how to model responsible, intelligent technology. Tools like ChatGPT are built on neural networks, a technique in AI that uses an interconnected, layered structure inspired by the human brain. The result is an ever-changing, human-like system that learns from mistakes and constantly improves.

“We’re experiencing a bit of a moral panic because we don’t understand the endpoint with AIs,” he says. “Will they become conscious or self-aware? We don’t know what consciousness is. We don’t know what self-awareness is. But the notion that AIs could achieve that is exciting and frightening.”

Childhood as a ‘1950s-Style Nerd’

Craig Newmark in high school. (Credit: Peter Vidor, left, yearbook photo, right.)

Understanding his own brain has taken Newmark decades. Growing up in suburban New Jersey, he says he could have used an AI to help him fit in. “In grammar school, I would want to show off what I knew and raising my hand to be called on,” he says. At one point, he thought this “dysfunction” was autism, but a casual conversation with a friend and behavioral expert confirmed it is most likely not.

“I would like to find an effective way to tell every parent in the species, ‘If your kid is doing that, tell them to stop.’ Why? All the other kids will hate you.” He hung out with a small group of like-minded kids and dressed like a “1950s-style nerd,” complete with a plastic pocket protector and taped black glasses (really).

In the future, Newmark expects children will spend more time online, but it won’t be like today’s post-and-view social media. Instead, they will “work together” in group environments like the metaverse, where they may benefit from strategically placed AI characters that can help teach them how to treat others. “I was a jerk, and no one told me that in a useful way and repeated it until it sunk in,” he says.

(Credit: Krongkaew / Getty Images)

Today, at least over video chat, it’s hard to imagine Newmark as a “jerk.” He has a calm presence and speaks in carefully worded sentences that regularly end with a story about a friend or family member. But his inability to “read the room” continued into adulthood, and he was particularly confrontational about technical matters. “People tried to tell me that was a bad idea, but it took over 20 years for that to sink in, which is not good performance,” he says.

Newmark has also experienced a fair share of antagonism toward him, particularly from critics who say Craigslist played a major role in weakening the journalism industry by taking revenue from print publications’ classified section. Newmark quickly challenges any implication that his site was the sole contributor, however, citing studies that found the switch to digital advertising models also played a role.

He expects AI will bring about the next significant change for journalism. “It may become more investigative—you have to go find out what’s going on out there, then feed that into an AI system and it writes it up for you.” (During our interview, he made a note to email the Craig Newmark Graduate School of Journalism (CUNY) to discuss incorporating AI tools into the curriculum.)

An AI-Assisted Future

In his day-to-day, Newmark says he would like an AI assistant to tell him when he needs to apologize, and to remind him of people’s names and how they met at the start of a conversation. Or, when to stop eating: “Stopping is important.” He also hopes AI can prevent subway delays, optimize traffic lights, and detect medical conditions faster than a human can.

One aspect of his life that’s unlikely to change with more technology: A love for pigeons, particularly one he named Ghost Faced Killer. “Sometimes Ghost Faced brings over all of his friends, and that’s too many,” he says, noting that his wife, Eileen Whelpley, had just gotten home and closed the window shades to deter them. “Seven pigeons is the max, we think. They see us in the window and waddle right over.”

Tweet