Key takeaways from AINEXTCon Seattle 2020: AI is eating the world.
It's been almost a decade since Marc Andreessen (also from the University of Illinois at Urbana–Champaign — in a bygone era) wrote an op-ed for the Wall Street Journal entitled Why Software is Eating the World.
No offense Marc — but as I absorb all that I learned at AINEXTCon Seattle 2020, it is time to note that AI is eating the world.
The recent Seattle-based AINEXTCon (actually held in nearby Bellevue WA) conference led with keynotes, and called out both the cutting edge advances in AI, and the very real ways in which AI-infused technologies are disrupting companies, both new and old.
Some of the highlights in the keynote addresses included:
Danny Lange — Unity gave examples of how gaming models can be used to generate large scale labeled data that can be used to train real-world ML models, such as for autonomous driving. Using virtual generated data could dramatically increase the training time to real-world pilots and implementations. Thought provoking indeed!
Joseph Sirosh — Compass (former VP AI at Microsoft and Amazon, now CTO at Compass) covered how the slow-moving world of real-estate is being transformed by AI. It was amazing to see how AI could help disrupt the traditional MLS-based industry that is currently in an intense battle, fueled by AI-based technologies, with Zillow, Trulia and others.
Erez Barak — Microsoft gave an overview of the lessons learned from Azure ML vis-a-vis customer patterns and practices. He outlined the breadth of skills involved in the transformation, from professionals in data analytics, MLops, and data science.
Bindu Reddy — Reality Engines gave a comprehensive overview of the history of AI all the way to the present day promise of autonomous ML.
Big gap remains between AI experimentation and live scaled AI deployment.
One of the more interesting validations from Erez Barak was the clear gap between attendees (tools providers and enterprises) who had started AI-driven trials, and those that were actually in the implementation phase.
In a quick “put-your-hands-up” audience survey from Erez with the hundreds in attendance, the audience results matched the reality today, as tracked by various market analyses— such as from BCG in a recent MIT article. AI deployments are just getting going, and less than 25% of projects and pilots have been deployed and adopted.
Source: RESHAPING BUSINESS WITH ARTIFICIAL INTELLIGENCE • MIT SLOAN MANAGEMENT REVIEW
Many different following breakouts sessions were comprehensive and covered a vast range of topics that ML is addressing, from autonomous vehicles, recommender engines, predictive maintenance, and deep fake detection.
With AI— it’s really hard to get from business strategy, to an AI project pilot, to a successful scaled implementation.
With all the ML frameworks, big-data analytics, and advances in cloud deployment one factor stood out: The rate of change in technology is so fast it can seem like you need an army of PhD.s just to track the changes!
The cutting edge advances require enormous compute resources. Just a couple of days Microsoft announced the Turing NLP Model that employs a staggering 17 Billion parameters. And in case you are wondering, I do not recommend trying to train this model on your 16GB MacBook Pro!
Talent remains scarce.
Talent is not only scarce, it is increasingly getting specialized.
So an engineer who understands predictive maintenance would be hard pressed to track the latest advances in NLP. Also, while all sorts of examples are available on github, applying the AI/ML techniques to your product requires considerable effort in acquiring & wrangling the data, select apt models, setting the optimal hyperparameters, and managing the deployment of the models.
Source XKCD : A hapless CEO trying to understand his ML Architect!
What to do next?
An unintentional result is that decision makers are left even more confused about what strategy to apply to create a compelling product and to defend their business advantage. If all of this technology is democratized is there a unique value to be established for your product?
We look forward to tracking the AI gap and offering real-world solutions for B2B tech companies and forward-looking enterprises looking to leverage the incredible potential of AI-powered technologies.
And most importantly, at WorksMachine, we look forward to helping you make the transition from strategy, to measured pilot, to scaled implementation.
Aloke Gupta, PhD
Partner, WorksMachine | Product & AI Strategy