We’re excited to bring Transform 2022 back in person on July 19 and virtually July 20 – August 3. Join AI and data leaders for insightful conversations and exciting networking opportunities. Learn more about Transform 2022
Are you ready to be a centaur?
What is the most important part of an AI system?
Is it the terabytes of data you use to train the base model? The billions of weights and biases sitting at the peak of your gradient tower? The carefully designed network architectures built on decades of brutal hard work?
What bottleneck are we in?
Is it that our GPUs are just not powerful enough? Do we need some smart architecture tweaks and a few points of accuracy to lead us to full automation? Or maybe shoving hundreds of millions of dollars into the money pit of data labels will click us into the future?
Or maybe, just maybe, we’ve been thinking about everything the wrong way.
We may have already entered a new paradigm for AI as a direct result of the rapid rise of deep learning techniques. Perhaps the most important part of your AI system is the person who operates it.
Power of the people
With all the focus on full automation and level 5 autonomy, it almost seems silly to focus on the person operating the system. After all, they are only temporary. However, as the adoption of AI within the enterprise accelerates, we see a very different picture emerging.
Overwhelmingly, the success of AI initiatives comes down to transparency, control, and trust. Eighty-four percent of enterprises still don’t trust AI, and a large gap in specialized AI talent — exacerbated by the pandemic — is one of the biggest barriers to AI adoption. The modern hunger for automation can’t wait for a workforce of hundreds of thousands of AI experts to fail.
All of this points to a critical need to rethink the way we build AI systems. How do we empower the citizen data scientist and bring the next tranche of AI users into the fold? We must stop seeing these as autonomous systems with incidental people. People are central to the development, use and maintenance of these systems.
Enter the centaur.
After Garry Kasparov’s famous loss to Deep Blue in 1997, the world watched with bated breath and wondered what the future would hold for people in chess. One person who did not wait was Garry Kasparov. In the truest expression of “If you can’t beat them, join them,” Garry Kasparov teamed up with a chess program called Fritz 5 to become the world’s first Centaur. In 1998, he took part in the world’s first centaur chess match against Veselin Topalov in combination with ChessBase 7.0.
Even today, with two decades of AI advancements under our belts, centaurs are competing with the best AI in the world. Given the apparent complexity of benchmarking centaurs with pure chess AI, the exact state of the art is somewhat controversial, but Garry Kasparov claimed in 2017 that there was “no doubt” that “one human combined with a set of programs is better than playing against simply the strongest computer program in chess.”
The paradox here is that human control and leadership add value even when the AI performs at levels that are clearly superhuman. The assumption that sufficiently advanced AI will eliminate the need for humans seems incorrect. Instead, we now have the task of creating the right interface for mutualism between us and AI.
Even massive organizations, steadfast in their commitment to artificial general intelligence, have begun to embrace more holistic approaches that recognize humans as a necessary part of the process. Obvious evidences include the increasing focus on “little shot” rather than “zero shot” learning, the closely related rise of prompt engineering, and Microsoft’s promotion of Machine Teaching.
Famously, OpenAI has even started incorporating humans into its training architecture effectively. In a recent paper, they dramatically outperformed state-of-the-art summaries by integrating a human feedback loop directly into the structure of their experiment. That’s a lot of human involvement for a field that is supposedly about automating people away.
But we should not be surprised.
History repeats itself and repeats itself
The first industrial revolution started with steam power and iron, but was built with the loom and machine tools. The breakthrough was critical, but the interface with people was what changed the world.
The second industrial revolution started with steel and sparks, but was built with the rail and the telegraph. Even when technology was crude by our modern standards, it made its way into the critical passageways of everyday life.
The fourth industrial revolution – the AI revolution – is underway. Data infrastructure, machine learning and cloud computing are important factors, but the core technology will only become clear in retrospect. What is clear is that the core interfaces that will echo through history have not yet been developed.
Despite all the incredible work we’ve done to improve technology, the vast majority of our AI interfaces have remained unchanged for decades. We are sorely lacking in the interfaces we need to enable self-driving cars, automated assistants and other theoretically groundbreaking technology.
This is our challenge. The next generation of AI problems should focus as much on user experience and human cognition as it does on developing and improving massive neural networks. We must learn from the lessons of the past to recognize that these human-machine interfaces we are building are not intermediate stages on the way to a utopian, automated future. They are the future.
So I ask again.
Are you ready to be a centaur?
Slater Victoroff is founder and CTO of Indico Data.
Welcome to the VentureBeat Community!
DataDecisionMakers is where experts, including the technical people who do data work, can share data-related insights and innovation.
If you want to read about the latest ideas and up-to-date information, best practices and the future of data and data technology, join us at DataDecisionMakers.
You might even consider contributing an article yourself!
Read more from DataDecisionMakers
This post Centaur soars: How a decades-old paradigm is changing the way top institutions look at AI
was original published at “https://venturebeat.com/2022/04/02/centaur-rising-how-a-decades-old-paradigm-is-changing-the-way-that-top-institutions-look-at-ai/”