In this recent thought-provoking piece in Aeon magazine Huw Price makes a plea that society as-a-whole must tackle the potential challenges that come with the growth of machine thinking and artificial intelligence (AI) systems. It’s just too risky and important to leave it all to Google and Apple (though he doesn’t name names like I have). He concludes:
Our grandchildren, or their grandchildren, are likely to be living in a different era, perhaps more Machinocene than Anthropocene. Our task is to make the best of this epochal transition, for them and the generations to follow.
We have plenty of dystopian (albeit fictional) visions of how not to manage this transition. The likes of Arthur C. Clarke, Philip K. Dick and Michael Crichton, amongst others, have found this to be a rich vein for scary visons of robots taking over, or at least challenging, humanity’s dominance of the planet. [I must admit I am rather enjoying the new, and very contemporary, take on the latter’s Westworld from HBO]. On the other hand, Isaac Asimov had AI coming the rescue of humanity in his classic Foundation series (albeit with human-induced hiccups and diversions along the way).
Apart from triggering memories of past sci-fi masters this article got me thinking about some fascinating, and incredibly promising, research, highlighted in two recent meetings which I attended. This research benefited from the benign and innovative use of ‘big data’ and networked learning to enable novel pathways to be explored which, to my view, would have been either too slow to deliver results and/or consumed too much resource, just a few short years back. It also triggered some novel, inter-disciplinary collaboration which could be a model for the future of research across the bio-medical sector.
The first examples were highlighted at a meeting at DeepMind health looking at patient and public engagement. You can see these examples presented in context and in full here. In this case, medical researchers benefited from working closely with DeepMind to elucidate issues with acute kidney injury and age-related macular degeneration. In both examples patient advocates and clinicians extolled the virtues of these collaborations, plus the enabling and positive role of DeepMind in this research. They also acknowledged the emerging issues around data governance regarding the use (and abuse) of patient data and input of patient views. It was interesting to me that this latter point triggered a lot of subsequent discussion, including this report on the BBC website. I am not naïve, unblinkingly trusting nor dismissive of issues around data use and governance. However, I do believe these debates are a distraction from what we should be celebrating i.e. genuine, well-meaning and care-directed, collaborative research – research which has led to real breakthroughs in diagnostics and patient care. I am sure there will be much more to follow, if we can overcome the data sharing or, more particularly, data abuse fears.
The other examples which struck me as relevant to this debate came out from a more recent meeting held just last week at Imperial College London. This Global Challenge Showcase brought together researchers from across a wide range faculties at Imperial to present ideas on how their research could impact on one of 3 major areas: Data Revolution, Engineering and Health and Well-being. Over one and a half days in their swanky new ThinkSpace office in White City researchers from Imperial presented some fascinating and promising research avenues and updates. Quite a number of these presenters showed concepts and ideas which rely on big data and/or processing power to drive and speed the research. Two strike me as relevant to the evolution of AI debate.
Dr James Rosindell showed how big data visualization can utilize fractals to store and present massive data sets in an innovative and accessible way. Just have a look at the truly stunning Tree of Life from OneZoom to see how this can transform digital data visualization. Cool and clever (and lots of potential future applications). The second example was much more applied and could be of massive interest to the pharma sector. Big pharma continues to struggle with shifting the drug development model away from slow, trial and error methodologies. Prof Michael Johnson demonstrated a superb proof of principal model using gene-regulatory network analysis (and lots of data crunching) to develop novel drugs. In this case the pipeline for the discovery of a potential new anti-epileptic agent was shortened to 2 years. One of background articles relating to this line of research is published here. It is hard to imagine this could have happened so quickly without systems level analytics, aided by data processing advances and new applications.
So that brings us right back to the original premise – how can we ensure the potential superior ‘thinking’ and processing power of AI is applied in the right direction (and no SkyNet emerges to wipe out us messy and imperfect humans!). I am convinced that the 4 examples above (plus many others going on in research labs across the world) point the way forward. Open collaboration, inter-disciplinary cooperation, innovative thinking and most of all public demonstration and explanation of the benefits of AI will enable useful and powerful positive applications to emerge.
Astrocyte Consulting are working to enable more and better inter-disciplinary, ‘de-siloed’ collaboration in bio-medical research. Please contact me if you need help in this arena or would like to discuss this topic in more depth firstname.lastname@example.org.