One of the most overhyped concepts in business today is AI (and machine learning). From murderous automated cars to covert surveillance, horror stories about AI are rife. But what is the truth about this revolutionary technology and how are businesses using it?
Today I talked to an expert in the field, who helped ground some of my more fantastical fears in the reality of AI’s largely benign omnipresence.
Steve Meier is the co-founder of Kungfu.ai, a consultancy which helps businesses find AI-driven opportunities by providing strategy and development services. Kungfu.ai is both a consultancy and AI developer, creating bespoke AI solutions in computer vision, language processing and predictive systems.
Meier has worked closely with IBM, developing Watson APIs to create unique sales, marketing, and event solutions. He studied at MIT Sloan School of Management, specializing in business AI, as well as the Luma Institute for Human-Centered Design.
Meier is convinced that AI can bring about a better world, and need not be seen as sinister, threatening, or impossible to control. Before we ask why he’s so sanguine, it’s important to define our terms of course since AI can mean something different to every speaker. For hardliners, AI are systems capable of passing the “Turing Test” (a blind test of AIs with the aim of fooling human interlocutors into thinking the AI is also human).
For Meier, AI is simply “something synthetic we have created that can do the things, and think the things, that otherwise we could do.”
The Robots aren’t Coming to Kill Us All
He downplays the more hysterical visions of AI, explaining that AIs are great at replacing very labor-intensive, manual, and repetitive tasks, freeing humans from having to do them. Data entry, photo recognition, production line activities and other automatable tasks can be readily handed over to specialist AIs. This frees up human workers to focus on the more skilled, interesting, and challenging jobs.
Meier sees the next frontier being the greater generalizability of AIs. Eventually, pre-trained AIs will be transferrable to new tasks, by means of machine learning. Inference will become a key area for development, so that an AI doesn’t have to be fed 1000s of examples to reach a decision.
Beneficial Effects of AI
Aside from the obvious incentive to use AI systems to advertise to increasingly specific markets, there are many other sectors benefitting from AI and machine learning. Growth in AI is anticipated in healthcare diagnostics, pharmaceutical research, robotics and manufacturing, education and more, with potentially hugely positive outcomes.
Meier has a project in healthcare now, working in collaboration with doctors and care providers to scour patient records to map relevant conditions and make connections between large volumes of data. Insurance is another pain point in healthcare which could readily be streamlined with AI.
Climate change opportunities include identifying the course of future hurricanes, so that proper preparations can be made. More sophisticated models can be built upon mountains of data. It’s already happening with SmartHome devices, making every home more fuel efficient.
The Problems with AI
Massive data centers are required to train AIs to understand language, and these leave an equally vast carbon footprint. Electric cars and blockchain technology are both bold new developments which eat up millions of electricity-hours in manufacturing or data mining.
However, things are improving as computational power becomes more cost-efficient, due to improvements in speed, and developments like transfer learning, where a model trained on one task can be transferred to a new but related task, giving it a kind of “head start”.
Blockchain and AI
Blockchain is widely misunderstood. Although it may have a huge carbon footprint, the bricks and mortar banking infrastructure its replacing is at least equally destructive. Efficiency savings in things like smart contracts and escrow payments are already visible.
When Blockchain and AI converge, there will be major opportunities, and Meier sees this as both inevitable and exciting.
Data Privacy and Transparency
Many of the concerns around AI center on how the data it feeds upon is being obtained and shared. Although there are differing degrees of transparency between the big data leveraging companies, biometric ID and 2FA at least provide us with better AI-driven data security.
Pandemic Response and Data Science
Kungfu.ai gives away some of its tech solutions to good causes and underprivileged communities. An example would be their work on COVID misinformation, helping track the sources of dangerously false material online.
“We are in a massive, underappreciated information war,” says Meier.
In a sense, AI is propagating the problem of misinformation, as algorithms drive more and more extreme content to users to maximize engagement. However, it can also be used to track and neutralize such misinformation. As with all technology, AI has no inherent morality, only the morality of those using it.
AI and Science Fiction
In terms of science fiction presentations of AI, Meier sees “Her” as being a more convincing story than “Terminator” or “Ex Machina”. We may become more emotionally dependent on our AI-systems. There are start-ups generating virtual versions of deceased loved ones using social media data scraping. This seems both potentially exciting and creepy at once.
However, Meier believes that General Intelligence is probably very far away, and AI becoming conscious or hyperintelligent is simply not a significant worry yet.
AI in the Here and Now
We already use AI tools daily — autocomplete on Gmail, voice queries to Siri and Alexa, Google Maps route planning and audio systems which filter out background noise are all in common use. There is no doubt that AI is here to stay, and Meier believes we need better education on what AI is, how it works, and the positive uses it can be put to.
So there’s no need to build that Terminator-proof, off-grid bunker in the hills just yet.