CAI has problems: Together we hold the solutions [1/3]
This is a three-part article about problems AI is introducing and how to fix them with blockchain’s distributed ledgers.
Part 1) To reduce loss of jobs, we can:
• Make AI more accessible (rather than proprietary)
• License our knowledge (rather than sell our heartbeats).
• Provide tools for learning new skills (not automate learning)
Part 2) To reduce loss of control, accountability and power imbalance:
• Provide open source tools we can all use (not consolidate in few companies)
• Give bots license plates (not inhumane machines with no reputation)
Part 3) To regain privacy, self-governance, and democratise AI, we can:
• Own our own data (not give it to others to publish).
• Build secure Conversational UIs (not lose trust in others).
This article explains solutions and organizations that are solving these problems.
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
Outlining AI, the reflection of ourselves
Artificial Intelligence now tells us where to drive (maps), where to eat (Yelp, OpenTable) what to do with our finances (WealthFront, NutMeg) and how to manage our health (FitBit, Apple Health). They tell us what the weather is doing, what we should wear and whom we should date. Talking to us in our homes, suggesting actions from our phones, wearables, appliances and cars, AI is surrounding us. They speak and recommend.
Analysts from Gartner to Forbes to Deloitte to McKinsey predict these system will offer immense economic and societal opportunities. There’s also a fear in the air as luminaries from Bill Gates to Stephen Hawking to Elon Musk predict AI as top societal concerns. It is as if these chittering children of automated manufacturing will emerge, like climate change, are surrounding us. AI makes guest appearances in our phones, speakers, cars, refrigerators. AI can beat us at our own games (namely Chess and Go). And some fearfully cry “The Singularity!” That famous scenario in which we are swallowed by our synthetic offspring. Whether reviewed as SkyNet, HAL 2000, West World, Ex Machina or perhaps some less somber scenario, the human vs AI conflict seems inevitable. The conflict seems to be starting in multiple fields, including work, control, privacy, and the very definition of humanity.
Part of this fear is because “AI” (like robot) is a poorly defined term. The word Artificial, etymologically derived from the same root as “art,” is a cultural construct of artifice. It’s an adjective which, at least until recently, has been more commonly associated with prosthetic limbs, colors, and flavors. The adjective has a positive ring. Meanwhile Intelligence comes in many, many forms. While often defined in terms of humans there are many forms of intelligence exhibited by Corvids, Cetaceans, plants, and Homo Sapiens. As far as I can tell “intelligence” commonly has to do with predicting, adapting and inventing. Despite being well researched it is, quite unfortunately, still measured against a human behavioral norm. And we definitely need to steer clear of mistakes made before the days of Copernicus: humans are not the center of the intelligent universe.
Top 3 Most Popular Ai Articles:
1. Intro to Deep Learning in Social Networks
Being clear about both terms, artificial and intelligence helps me understand what it is: A social construct, perhaps a myth.
AI isn’t an emerging, intelligent species: it’s a group of technologies people made. AI is an implementation of multiple things: input / output methods of sounds and images, storing, processing and predicting patterns that allow social and physical navigation, etc. For most people that group of technologies is unknown. And that introduces doubt, uncertainty and sometimes fear.
“What can it do better than me?” they think, “Can it think?”
We built it therefore we can understand it and we can interface with it.
Some of these systems, specifically the interface to AI, are known as bots.
BOT = UI | BOT ≠ AI
Bots, an interface to AI, can be broken into taxonomic categories. Bots can be scripts that monitor traffic, software that methodically launches attacks, and algorithms that converse. Bots can be chatbots that read, then write back. Messaging bots that dwell in the apps we use to message one another. TwitterBots, SlackBots, KikBots, and others are all relying on text as the user interface to the information it provides.
Let’s now add ears and a mouth to our bot so that it can speak and listen. If we add audio the bot becomes multi-modal. This bot (like an Amazon Alexa, Echo or Dot) is also known as an Assistant, sometimes called “smart speakers. It has a voice recognition capability that converts spoken words into written words. When you speak to the system and say “Star” it sends that audio recording to a computer that then looks for the best match between the sound and the letters. It might come across some similar-sounding word, like “scar,” but regardless of the results this ability is based on past data that was collected. Lots and lots of people mapped the sound of a word to the letters of the same word. This, ultimately, gives the system ears and a mouth.
When we add eyes to our bot we start to see the future of bots emerging. Just as you, dear reader, have spoken to someone via videochat bots can now do the same thing. They can see you via computer vision (CV). They can identify objects that are in the background. They can measure your appearance. They can also, by having a face and waving their hands or shrugging their shoulders, express themselves better. These bots may now be deployed in real-time 3D systems such as VR and AR. Conversational avatars may also appear in videochat channels, such as Skype, Messenger, Signal and others.
Whether we use text, voice, or sign language we’re still having a conversation. Whether a chatbot, an assistant, or a video chatbot, these are conversational user interfaces, or CUIs.
CUIs are my own obsession and the raison d’être for our company, Botanic Technologies. Some CUIs use, as we humans, faces and hands to converse. At Botanic Technologies we have implemented these multi-modal systems for nearly a decade.
Bots are an interface to AI. This means that bots are also an interface to the problems that AI presents.
Problem #1: Displacement of Jobs
A fear of losing jobs to AI is a top stressor for many people, and Americans in particular. According to a report from Udemy in June of 2017, 43% of American workers link their stress to a fear of losing their job to artificial intelligence. This fear is confirmed. Reports by both Pew Research & Oxford University indicate that AI will impact at least 43% of jobs by 2025. Globally the numbers are higher.
This is a concern more evident than the numbers.
Most of us work by selling our heartbeats. In order to make a living almost all of us have an hourly wage or an annual salary: the money that goes into your pocket is proportional to the heartbeats you put into your job. Especially true in unskilled and manual labor, human heartbeats are being bought and sold as the primary measurement of value.
Different people make different rates for their heartbeats — some people’s heartbeats are more valuable than others — so knowledge makes heartbeats more valuable. The knowledge multiplied by heartbeats is what allows knowledge workers to make more than unskilled workers.
AI does not operate on this same formula. AI can accumulate knowledge incredibly fast (and it doesn’t have a heart). For example, as the Guardian states it, “AlphaGo Zero took just three days to master the ancient Chinese board game of Go … with no human help.” The interesting thing about AlphaGo Zero is that it used Tabula Rasa learning methods and in a matter of weeks learned all the games of Go that people have been playing, adopted many of the same playing patterns, used some, discarded others, and invented new strategies that people have never used. Based on John Locke’s philosophy, Tabula Rasa was the theory that we’re all born with a “blank slate” and that knowledge is accumulated by our sensory experiences. In a similar way AlphaGo Zero is not only accumulating knowledge on that blank slate faster, it’s actually outstripping the current body of knowledge and discovering new ways to play. This is important because the Tabula Rasa method, also called reinforcement learning, can be abstracted from Go and used in other systems with contextual parameters. So people are afraid of this because it shows that AI can learn really, really fast and about new things we don’t know. In a heartbeat.
Second, after the initial cost of building out the system, AI is able to access, process, and automate many types of information faster and less expensively than people. These information management skills range from natural language summarization of sports stories to industrial machinery prediction and these AI systems far outpace humans in their speed both accessing and processing information. There is a concern for the potential loss of jobs, especially for knowledge workers, and that the resultant outcome will be a cumulative effect.
One important cumulative effect is that AI will generate more wealth in the hands of those that least need it. This kind of monopoly control of markets (and therefore capital) will dominate some parts of labor and push down many wages (for those that keep their old jobs). This then will domino into more AI deployments to replace more people, causing economic collapses, wars, revolutions, and other exciting events. We can see the beginning of this with large automated systems like Uber, in which the backend of the system is consolidating wealth as the drivers each feed more data into the system to provide autonomous navigation skills.
Some jobs will disappear and others will appear. Automated manufacturing has always generated jobs and this fourth industrial revolution won’t be an exception. But still, how do we mitigate losses?
Solution #1.A: Build conversational interfaces to AI.
Because they are familiar conversational interfaces are easily adopted. Familiar interfaces, like the GUI or, now, the CUI, allow simple access to complex systems.
Once upon a time, the calculator was a desk job. When digital calculators were invented those people’s jobs were displaced. Today we all use calculators. Once upon a time the Computer was a desk job as well. Today they are a staple part of the work we do. You are probably reading this on a computer, such as a smartphone or laptop. New jobs appeared as computers became more mainstream. The same will happen with AI, provided we preserve simple interfaces to these systems. Bots now function as tax accountants. This means the CPA is now going the way of the Calculator. And the CPA of tomorrow will have a simple, conversational interface much like the calculator of yesterday had a simple, graphical interface.
AI’s rise won’t necessarily cause permanent, disastrous pain. Augmented intelligence, an extension of you gives possibilities for prediction, review, confirmation, learning, knowledge, dialogue, analysis and thousands of other uses available to you and expanding exponentially with only a temporary disruptive effect on the workforce.
Many jobs actually need to be replaced by bots. Customer-relation call centers have roboticized people for the last decade, forcing lower-wage workers to repeat, line-by-line, the words that appear as they navigate a conversation tree. Insurance centers that pay people to follow a prescripted path of logic to derive a calculation as they interview someone is mind-numbingly slow work that is an insult to intelligence. There are many jobs that, like a calculator, don’t require a human mind and may, in fact, be better if the human is removed from the equation. Financial qualification processes, retirement centers, government application centers, even the vast majority of job application processes have roboticized people. Many jobs need replacing by automation. What today look like if, because of concerns of loss of jobs to automation in the mid-1900s, people were still sitting at desks computing and calculating?
Jobs that contain drudgery, boredom, repetition and work within tight legal or technical compliancy restrictions are not jobs that are well-adapted to human laborers. Anti-money laundering regulation, for example, requires reviewing thousands of records per day and checking for tiny inconsistencies that humans just aren’t good at — and we get quickly bored doing that stuff. Research has shown that AI’s best-suited to predictable tasks when errors are cheap. As jobs becomes more complex and less predictable AI makes mistakes, and the more complex the job the more costly the mistake. With retraining and reskilling we can provide value for people missing out. Retraining and reskilling takes time, creativity, collaboration and tools. More on that in a minute.
With social interfaces to AI via bots and CUIs we not only keep people in the loop, we augment their skills and build on the very human traits such as creativity, symbolic recognition, and social interaction.
Solution #1.B: License our knowledge rather than sell our heartbeats.
As any Silicon Valley investor will tell you — services is a poor business model for a company. So why, as individuals, are we working this way?
How can we, as individuals, provide a product we each author? What if you could develop a bot that allowed you to receive an ongoing royalty for valuable things like your daily activities, social graph relations, knowledge, language, understanding, skills, or personal information that someone else finds valuable? Companies like Facebook and Google, companies that sell our personal data, already understand this and we can see the value of our data as measured by the size of these companies. Your data is the most valuable commodity in the world. As Derek Powazek once put it, “If you’re not paying for the product, you are the product.”
Software companies scale the value of their data by licensing it to others. It is time that users be able to scale the value of their data in the same way. We need to find scalable business models for individual knowledge workers. AI won’t take your job if your job is to inform what the AI knows. Computers and Calculators of the 1950s — the humans that computed and calculated at desk jobs — learned this lesson decades ago. They’re now programmers and many of them license their knowledge (or, more accurately, their companies license their knowledge). They make knowledge that can be licensed and some of them program AI. The challenge is to build bots that allow everyone to license their knowledge and build it in a way that is open, monetized, scalable, and free.
Q: How can people individually license their knowledge?
A: An Open Source Bot Economy.
Seed Vault, a Singaporean company that is providing a blockchain-based solution, has begun building this ambitious solution. The SEED token is an open-source solution that enables a bot economy. Like Wikipedia or Linux people contribute knowledge. Unlike Wikipedia or Linux contributors are compensated for the data they each contribute. This allows us to establish an interface to AI that is a democratic, trusted, and fair. Companies, individuals, groups and bots may exchange value equally. This is done via CUIs. It can solve many of the problems of job loss from automation by allowing people to contribute to an economy in which they can be paid by improving automated systems in the same way that a software programmer improves a computer. And other projects are taking complementary approaches. Open Ocean, confronts large data. BotChain is about scripted bots.
This is an inflection point at which, for the last several decades we have trained people to behave like robots, at call centers, in service roles, and in many knowledge worker roles. It’s time to allow machines to take on the more automated tasks, allowing people to fill in automated processes with more thoughtful approaches. It is the graduation from calculator to programmer.
We must squarely address the changes that confront us. Emerging technologies, like blockchain, can solve many of the problems that AI introduces. The problems AI introduces are problems we have invented, and that means there are solutions we can build.
The Convergence Ecosystem sees data captured by the Internet of Things, managed by blockchains, automated by artificial intelligence, and all incentivised using crypto-tokens. The Convergence Ecosystem — open-source, distributed, decentralised, automated and tokenised — is an economic and societal paradigm shift.
To be continued . . .
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
Thanks: Massive thanks to the formidable Michael Tjalve, of Microsoft’s Bots for Good for his input, critical thinking, informative examples, and philosophic temper. Thanks also to Lawrence Lundy of Outlier Ventures, and Ben Koppleman of SEED for their reviews, additions, and input.
