Last Wednesday, Blockchain Hub hosted another meetup at York University. This one focused on two technologies that are both leading-edge, both on the threshold of changing the world, both independent for now. The first, of course is cryptocurrency and blockchain technology. The second is artificial intelligence, or AI. The bulk of the presentations focused on the latter, as many of us don't know as much about AI as we do blockchain.
As per usual, the meetup started at 6:30 PM but the presentations didn't get rolling until 7 PM. The half an hour in between was reserved for "networking," but a lot of us just took our seats and waited.
Part of the pre-presentation featured Othalia Doe-Bruce presenting information on Blockchain Hub's course offerings and virtual incubator service.
These services are part of Blockchain Hub's aim to become a startup incubator and accelerator in the cryptocurrency world, in partnership with York University's Lassonde School of Engineering. In what might have been a first, this meetup was llivestreamed on their Facebook page.
Phillipe Chevry: Emergence and Convergence
The first presenter was Phillipe Chevrey, who streamed his talk live from Montreal. He's the guy on the right in the big split screen:
In his career path, he started off with gaming and then moved into machine learning. After discovering Bitcoin, he explored merging machine learning with blockchain technology. He observed that there was a difference in the emergence of the two technologies: cryptocurrency, being at heart an underground movement, owes a lot to self-learning and a tinkerers' ethic. (He didn't mention this, but the steam engine grew into maturity in the same way.) AI, in contrast, grew largely out of the academy and is deeply dependent upon academic development. In contrast to AI, academics are coming into the blockchain space relatively late. (One exception to this is ZCash, whose core dev team was dominated by professors of cryptography.) That's why, in contrast to AI, a lot of research and development in the blockchain space is being undertaken by startups.
With the latter, the focus is going to be how to decentralize the very complex calculations that AI needs. We all know that a large distributed network of ordinary computers can swing enough computational power to rival a supercomputer. (Just think of how much crypto those malicious botnet operators have made simply by leeching surplus computational cycles from their networks of victims.) A blockchain can validate resource use and the associated cryptocurrency can incentivize suppliers of memory and computing power. The startups converging blockchain and AI will tinker around with incentive frameworks as part of their research and experimentation.
Based upon what he said, startups and their hard-charging guerrilla-building culture will be at the forefront of converging blockchain with AI.
Lana Novikova: What's In A Word?
The next presenter was Lana Novikova, the founder and CEO of Heartbeat AI.
Her company uses AI to attach emotional meanings to words, in order to gauge people's emotions on certain subjects. In contrast to the usual structured-survey method, her raw data comes from open-ended survey questions like: "Do you feel AI will be important to the Canadian economy?" This question was one of five she gave to the audience as a live demonstration of Heartbeat's tech. After getting her responses, she revealed that the audience were more optimistic about AI, and less afraid of it, than a sample of ordinary Canadians. Nevertheless, the audience sample was still cautious about AI. We-all were trusting but not joyous, and were less emotional about it than ordinary folks.
The root method of Heartbeat is to partition words into emotional groups, like joy, fear or anger, and then break down each partition into more subtle sugroupings. Novikova herself used to be a clinical psychologist; Heartbeat has three to five trained psychologists to aid the AI machine to tease out emotions from words and also to check and correct for biases. One of her success stories was Heartbeat's prediction that Donald Trump would win the swing states he won. Funny enough, the level of dislike for Trump was in line with the dislike for Clinton. What counted was the type of dislike. The folks that disliked Hillary were more prone to feel righteous resentment and other emotions of dislike that are actionable. Novikova said that Heartbeat's predictions implied that Trump's victory was caused more by voting against Clinton than for Trump. In other words, it's more accurate to say that Hillary lost the election than to say the Donald won the election.
After the demos, Novikova moved on to the potential of emotional AI technology. Using the analogy of chess champ Garry Kasparov and Big Blue, she said that emotional AI would be unleashed most powerfully in partnership with real humans rather than as competition for humans. Artificial is a way station to augmented, to humans using emotional AI to give a boost to their emotional intelligences. (Just like, to use an analogy up my roadway, a motor plow has augmented human muscle but has not replaced the good old snow shovel. A plow-only job is quick but coarse; a snow-shovel job is slow but granular. Consequently, many professional crews comprise a truck plow and one or two fellows with hand shovels.)
One practical result we can expect from augmented emotional intelligence is much better career locators. The benefit: less floundering around in the job market until something clicks. Another practical, though softer, result would be AI-powered informal psychotherapy and companion services. This tech would help depressed people get back on their feet. As an example, she adduced the Scarlett Johansson movie Her.
Alpha Block: Helping To Face Complexity
The next presenter was the head of Alpha Block Capital, involved in applying AI to financial markets, finance itself and tech.
He talked about Alpha Block tackling complexity, which does include spots of chaos and error. Blockchain technology does not address this space because it was not designed to; its aim was to decentralize the monetary system and move it from single-point-of-failure to peer-to-peer. Alpha Block aims to take data sets from the blockchain and merge them with data sets from the complex network of the World Wide Web. This should lead to connectivity between the two.
Cyptocurrency is revolutionary, but it's stuck in the old in a certain way. A blockchain network is human-made, though sustained by its coding, and is also linear and unidirectional. It's computation-centric, market-cap-centric [as we all know] and the initial distribution of the coins has a large effect. (If you were in crypto long enough ago to remember the "mining wars," in which miners routinely slammed Proof-of-Stake coins, this last point will stick.)
To sum up: blockchain networks, though revolutionary, could benefit from further innovation that would make their networking more sophisticated: for example, by attaching AI tech to a node.
One Ledger To Interface With Them All
The final presenter was David Cao, the founder of One Ledger.
Less philosophical, Mr. Cao came from IBM's enterprise computing division and discovered Bitcoin about two years ago. His goal is to tackle the tricky problem of cross-chain access, including cross-chain transactions.* True to his background, he's developing One Ledger as an enterprise solution to connect independent blockchains.
Doing so would provide interoperability to the large set of disparate cryptocurrencies. His tech offers a high-level interface of modules to the end user, backed by a low-level smart contract. The master contract will provide the cross-chain functionality, and will interface from above with the One Ledger network; this network will interface through smart contracts to each supported cryptocurrencies' blockchains. Cao plans to work in AI modules for self-learning, improvement and optimization.
*: There has been one guy, Ian Knowles, who's been working on this problem since at least 2013. If you're a fan of cryptocurrency trivia, the first atomic cross-chain transaction was an exchange of Qora for BURST in May 2015. Knowles' CIYAM provided the underlying tech for the transaction. Sadly, Qora faded into the night of the '15-16 altcoin bear market.
Panel Discussion, Mostly Philosophical
The final part of the meetup featured Othalia Doe-Bruce and the three live panelists for a Q & A session. Ms. Doe-Bruce asked some questions, like asking about how Heartbeat AI would be integrated with Open Ledger, but most of the questions were from the audience.
And there were a lot of those. The number of questions were well into the double digits, and there were more hands raised than questions taken. By this metric, the audience was deeply engaged.
Most of the audience questions were philosophical in nature. Since AI is still exotic, with deep knowledge confined to professors, academics and pros who are specializing in it, the philosophical nature of many of the questions wasn't unexpected. There were a couple of questions about the ethics of using AI to analyze human emotions. There was even one which broached the Singularity.
After a string of high-abstract questions, I managed to squeeze one in about using Heartbeat's technology to gauge sentiment on cryptocurrency markets to refine the good old Theory of Contrary Opinion. (To borrow Warren Buffett's words, be greedy when everyone is fearful and vice-versa .) Although my question did break the philosophic spell, she intimated that sentiment analysis of this sort wasn't on her list of priorities.
One subject was raised by the techies amongst the questioners: how do you integrate an irreversible blockchain ledger with dynamic and changeable AI? It was a pertinent question, given the centrality of irreversibility to the blockchain and the need for exactness. (Famously, Satoshi Nakamoto coded Bitcoin balances as 64-bit integers to eliminate the occasional inaccuracies introduced by fractionated doubles.) Consistent with his startup, David Cao said that it could be done with a new ledger on top.
Overall, this was a heady meetup. The subject was popular and engaging, and the speakers certainly held our attention. There are legitimate questions about how well AI can be integrated into blockchain tech. For now, the only low-hanging fruit seems to be tapping a crypto ecosystem for assembling distributed networking: using a blockchain like Ethereum's to record, keep track of and incentivize a system of hiring and paying folks for supplying their computers for AI calculations. Over and above that use case, we're looking at something of an unknown territory.
So there will be special bumps along the merge parts of the converging roads. But those roads will eventually be interoperably converged: these days, that's the way to bet.