We are excited to bring Transform 2022 back in-person July 19 and virtually July 20 - 28. Join AI and data leaders for insightful talks and exciting networking opportunities. Register today!
Somewhere between CUBE Tech Fair in Berlin and Next Block conference in Kyiv, I grabbed a rare opportunity to drink coffee and talk about artificial intelligence with SingularityNET’s Dr. Ben Goertzel. It might have been the most brain-jolting caffeine shot I’ve ever had.
There are questions that need to be answered with regard to AI’s role in the future of technology — and humanity — and if there’s anyone who might be able to shed light on that future, Goertzel is a good candidate.
Born in 1966 and with a PhD in Mathematics from Temple University, Goertzel is not only the CEO and chief scientist at SingularityNET, he is also chair of both the OpenCog Foundation and the Artificial General Intelligence Society. In his own words, Goertzel is “focused on creating benevolent superhuman artificial general intelligence and applying AI to areas like financial prediction, bioinformatics, robotics, and gaming.”
SingularityNET is the vehicle with which Goertzel hopes to achieve that lofty goal. A sort of “GitHub for AI,” although it is much more than that.
The company immediately garnered headlines when its token sale hit 400 percent oversubscription. But it might be best known for powering Hanson Robotics robot Sophia, and it is also becoming instrumental in funding other AI projects. SingularityNET and AI Decentralized recently announced the Decentralized AI Alliance (DAIA), an open industry alliance that is helping foster the development of decentralized AI technologies.
When we sat down to talk, Goertzel didn’t mince words. His views on where AI is heading are thoughtful but pointed.
“What you have now is a complete oligopoly of data and AI processing power,” Goertzel said. “That’s just the natural dynamics of capitalist economies, unless you build something to actively resist against that. That requires some subtlety. The cryptocurrency market has a natural tendency toward this model, but when you look closely at that, Bitcoin and Ethereum are oligopolies too. There’s a small number of large token holders who control those markets, and [the markets are] heavily manipulated.”
We’ll get back to the topic of democracy in AI later. When Goertzel and I first sat and ordered our personal choice of caffeinated beverage, the issue of the day was still Cambridge Analytica and Facebook. I wondered how Goertzel saw that whole debacle from an AI and machine learning perspective.
“There’s been a backlash, but not that big, because everyone is still using Facebook,” Goertzel said. “But people are grumbling about it. In China, they don’t care. They’re not up in arms about it at all. ‘They’re taking my data and using it to program our brains!’ It’s just normal. But in the West, people are unhappy. Of course, these things can shift. China is prone to rapid shifts.”
That opened up a discussion about the broader implications of AI, artificial general intelligence (AGI), and where this is all going. For example, how can we guard against the improper use of AI and data?
“The way I’m thinking is you need to build an infrastructure and a platform in which AI and data are decentralized, self-organizing, and participating in an interactive way with everyone on the planet,” Goertzel said. “If you can build a platform like that, then that’s quite resilient and adaptive with respect to whatever happens on the part of big companies and governments. Then whoever builds the next Facebook will have a decentralized fabric of data and AI to use as the backend for it.”
Goertzel is very clear, however, about where his intentions lie.
“I don’t want to build the next Facebook,” he said. “I do want to build the decentralized AI platform, and a decentralized data collection and management system, which means that a democratic Facebook-like solution would be a very thin layer on top of that.”
And the way in which the platform is built supports Goertzel’s assertions that oligopolies in AI and data management need to be avoided for the marketplace to succeed.
“When we designed SingularityNET, we specifically designed the economic logic to avoid [giving] a small number of parties the ability to take control of it,” Goertzel said.
While a platform that brings AI and data together in a decentralized, community-driven way sounds like an entirely achievable goal, the complexity of collecting, storing, normalizing, and managing the data portion alone is a significant challenge.
“That’s easier said than done,” Goertzel noted. “The last few months, we’ve been gearing up to do exactly that. We’ve hired a big AI team, and we’ve been thinking hard about how to architect a system that can do exactly that, this year and not in five years’ time.”
That’s a significant challenge. By building a platform that takes in everyone else’s AI projects and associated data sources, you’re attempting to create one of the most elastic data stores we’ve ever known.
“It’s partly about creating a framework that the participants can use to resolve these challenges,” Goertzel said. “We have an API of APIs, essentially, We could define right now what an AI agent for natural language processing (NLP) looks like, or an agent for computer vision, but it could be different in six months from now. We don’t want to be the expert committee defining how these have to talk to each other.”
The answer, in SingularityNET’s mind, is to allow the ability for the community as a whole to ensure that every AI is appropriately incorporated and make sure everything works in concert.
“We’ll put out the first approximation of these AI mechanisms to seed the platform,” Goertzel said. “But we also need to make a mechanism so that everyone that has an AI on the network can collaborate and vote on what the communication details should be for NLP, or computer vision, or reasoning, or whatever it might be.”
Those approximations? Well, let’s say they are hardly minor. While Goertzel knows he has to seed the platform with AI for all significant needs, he’s already working on algorithms and data sources that are significantly more interesting than those occupying the average data scientist.
“We’re putting up a set of tools that are working with DNA data,” Goertzel said. “We’re working with Project Shivom, which allows people to upload and store their genetic records on the blockchain. The people still own their DNA data, and they receive tokens for its use within research, so we’re creating the analytical tools for that information.”
That’s not the only advanced implementation of AI SingularityNET is working on. Languages are inherently crucial to global communication, so naturally the team is working on solutions in various areas of linguistics.
“We’re doing our own AI for language learning so that the AI can learn syntax and semantics from streams of text, and we’re experimenting with that using social messaging,” Goertzel said. “Right now, nobody can analyze Twitter in terms of the meaning of a tweet. AI can parse the tweet and attempt to gather keywords, but they don’t try to gather the real meaning of the message because it’s not grammatical. The AI we’re developing can parse the meaning of a tweet by learning the grammar of the tweet in various languages.”
Sentiment analysis on tweets and other social media messages is, frankly, worthless most of the time. The “burger paradox” often comes into play. “That burger was bad” and “that was a bad burger,” where the former uses “bad” to mean “good,” causes issues for sentiment analysis tools. Irony, sarcasm, metaphors, and proverbs trip these systems up all the time.
“It depends on the domain,” Goertzel said. “If you’re doing sentiment analysis in Bloomberg press releases on stocks, for example, it works pretty well. There’s not a lot of irony in financial news headlines! ‘Apple stock went up today — oh, just kidding!’ Sentiment analysis, however, is a very crude tool. It doesn’t tell you that much. If you want to go deeper, you need to know why people are saying these things, what is the context, and what is the root cause of these reactions.”
Goertzel explains the difference using everyone’s favorite TV show about AI and robots.
“It’s easy to work out whether people are enjoying the latest episode of Westworld or not,” Goertzel said. “What’s harder is working out whether the audience is sympathizing with the robots or the humans. There you’re getting into things like sarcasm and subtlety of expression. While we’re not building that right now, we’re putting together the framework that would allow someone to build that.”
Moving on from the uses of AI and the platform Goertzel is building to bring it all together, we started to discuss how AI and, ultimately, AGI might play out in the future. Who would control it, and how far away are we from AGI playing an essential part in our daily lives?
“I think the choices we have are the analysis of everyone’s data and the development of ultimately superhuman AI being in the hands of future massive corporations that are acting in league with various government agencies or, on the alternative hand, the world’s data and the world’s AI processing power being developed in a participatory way by a broad sampling of people around the world,” Goertzel said. “I don’t see an alternative to those two possibilities.”
While other possibilities have been discussed in various consortiums and throughout the past year at AI conferences across the globe, Goertzel may have a point.
“I know people like to fantasize about picking 100 noble, good-hearted geniuses, then [locking] them in a basement on an island to develop one superhuman thinking machine — the benevolent AI dictator,” Goertzel said. “But the track record of these things in human history is not good. We’re better off with humanity, warts and all.”
So the answer to AI and AGI growing benevolently alongside humans lies in our inherent ability to create and foster communities, Goertzel believes. That’s why he believes in democratizing and decentralizing AI and data management, rather than relying on a corporate and government-based, or centralized, alternative.
“A lot of this has got to be about creating community,” Goertzel said. “As a technologist, I’m focused on building the platform and putting the smartest possible AI software in that platform so that we learn, generalize, and reason. On the other hand, we’re small potatoes in the AI world. If we’re going to dominate, we have to build a humongous community to help them and us.”
While SingularityNET is working on the code and the platform, it is spending as much time building that community in the hope that a democratic AI and data management platform will ensure we all benefit from, rather than become targeted by, our future chip-based overlords.
The ultimate goal?
“Democratically created and guided AGI,” Goertzel said. “We want to foster the creation of AGI, but in a way that’s guided by everyone on the planet. AGI has tremendous potential for good, and the only way I can see that happening is if everyone takes part in making sure it is used for good.”
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn more about membership.