The last few days there was a considerable buzz around the conversation between Mark Zuckerberg (Facebook) and Yuval Noah Harari (author of Sapiens, Homo Deus and 21 Lessons for the 21th Century).
Here is the video (90 minutes):
Overall I feel and think that Harari came about in a more positive way relative to Zuckerberg on most topics. Zuckerberg is understandably protecting Facebook almost all the time due to his role as CEO, Harari is independent.
I invite all viewers to watch the body language of Mark Zuckerberg around sensitive topics related to Facebook and/or USA data policies.
Harari clearly is not a technology expert like Zuckerberg, he lacks skin in the game in this domain. In his book Homo Deus he shows in my view that his ‘Dataism’ is overblown. Human consciousness and existence is not just data. This is a fallacy. Consciousness is scientifically proven to be driven by not only our brains but also our bodies and even the universe itself (muons). So AI is very different from Dataism and Consciousness (see Roger Penrose for more depth). On top, recently George Dyson confirmed in an Edge.org post that we move away from digitization back to analog (quantum) systems driven by nature itself due to its control, context, complexity and performance. Even tough Harari is not a tech expert he sometimes does raise some interesting questions as shown below.
In this post I would like to review some of the key topics and questions being discussed by Harari and Zuckerberg and give my own questions and thoughts in this respect.
Are we more connected or not historically?
We are more connected than 10.000 years ago. Connecting people sometimes lead to more conflicts compared to being isolated says Harari. There is difference between being connected and being harmonized. I believe he is right looking at todays increasing conflicts in the world. People are more connected (virtually) than ever before but we also see some disturbing statistics around the world related to social media. Youngsters increasingly are lonely, depressed, anxious, have auto-immune diseases and burn-outs. It is not only driven by social media but empirical scientific research confirms there is a role here. So people are more disconnected with themselves than in the past and probably as a result less harmonized with each other as well and social media play a role in this. Checking in with yourself is a prerequisite to checking in with other people in an effective and authentic way.
There was a related dialogue around the value of offline vs online communities. Zuckerberg defended the online communities while Harari told about the value of offline communities. Both had a good argument here in my view but I would prefer offline relationships as a stabilizing force in an exponentially changing world. The shift should be less online time spend and more offline to avoid insanity as emerging mental health disease going forward and to create more harmonious societies locally. In short, Facebook is like Heineken beers: do not use it too much, it is bad for you. This was even confirmed by Zuckerberg himself when he talked about video virality in this dialogue. Imagine, compromising DeepFake videos being used on Facebook like in the Russian-Cambridge Analytica campaigns….
See also this highly successful and viral TED video by Carole Cadwalladr for additional Facebook criticism focusing on the Brexit and democracies:
Is technology (a.o. AI and social networks) favoring dictators / centralized systems or decentralized systems?
Historically, decentralized and distributed systems are normally and in general more resilient, efficient, innovative and effective relative to centralized systems (see also the great books by Nassim Taleb with Antifragile and Kevin Kelly with Out of Control for more research). The question is if this is still the case due to the nature of AI today. Why? AI today definitely seems to favor centralized solutions and systems in advancing nation states. E.g. China with Alibaba, Baidu and Tencent. The more data you have (in general), the better the AI or algorithms can perform and predict over time. This is Harari’s point too. AI is avoiding the information processing problems of centralized systems in the past. Valid point in my view.
However, there are a few new interesting developments related to AI which could impact Harari’s AI observation.
First, in the EU there is the GDPR privacy law which blocks the integration of data sources and as a result lowers the centralization power of AI, at least for now in the EU. However, this might backfire in terms of the competitiveness of EU relative to the US and more importantly China. Long term, GDPR might stimulate Trust in societies but it is all about the trade-offs we are willing to make in terms of personalization vs privacy as well as security vs privacy. The individual vs collective rights and benefits.
Second, the most advanced form of AI today is deep learning, a bottom-up connectist view on AI using advanced statistics and probabilities. Recently, there seems to be a limit to the performance improvements due to more data being fed into these deep learning solutions. The top-down, symbolic school of AI (dealing with semantics, taxonomies and real understanding of concepts) is getting more investments again to boost AI performance again. A new synthesis of bottom-up and top-down schools of AI is the cutting edge today globally in terms of performance. The good news is that the top down schools of AI do not necessarily require the same scale and huge data sources as in deep learning but clever (crowdsourced) startups which is encouraging for the AI decentralization movement.
Third, during the last few years in China we have seen significant progress in (photonics based) Quantum AI or Quantum Algorithms which amplifies the (simulation and optimization) capabilities of AI relative to AI written for classic supercomputers. Quantum AI favors centralization even more as it seems today in terms of better performance.
Fourth, MIT recently announced via Alex Pentland their ‘Federated Data’ initiative which allows for AI to integrate data from different databases without owning the data. This lowers the centralization power of AI.
In sum, on a net basis there are reasons to believe the centralization trend of AI might be mitigated in the future.
Moreover, it is a secondary discussion in my view. The discussion is not so much about centralization vs decentralization but more about the intentions of the humans driving this at the national levels. The long arch of history after the hunter gatherer societies tends towards more democratization and decentralization supported by different technologies. This is not set in stone but there is an evolutionary tendency. All technologies (linear, exponential and recently even quantum) can be used for both good and bad, for democratization (empowerment, more choices) and controlling (less freedom). Kevin Kelly talks in his seminal masterpiece What Technology Wants about the evolution of technologies in general towards more possibilities, options and choices. So AI is more a derivative of human intentions and less about its nature (centralizing or decentralizing).
Are social networks fragmenting societies and/or offering personalization of content, ads and services? (respectively are they bad or good?)
Zuckerberg claims people on average want personalization. Fragmentation is seen in filter bubbles and polarization. However, my take is that all real and deep personal and professional growth is due to out of the comfort zone experiences and outside our filter bubbles. That’s how you also can find your (new) passions and purpose(s). We don’t always know what we need. This goes beyond the discussion of personalization vs fragmentation. I believe social networks in the latter sense are both good and bad but bad in terms of inhibiting finding your passion and purpose in life.
Social media data location, storage and protection? Where?
Zuckerberg defends USA as centralized data storage compared to rogue nations. However, USA is just as bad as rogue nations in some aspects. I refer to the PRISM, Snowden and Assange. Zuckerberg is either very naive here and playing dumb.
AI and inequality:
Zuckerberg avoids this discussion in my view. This is too bad. AI will have a significant impact on (un)employment in the next 10 years globally. In the Netherlands (my home spot) for example, 1,3 million current jobs will be impacted by AI while 1,4 million new jobs (quite a few in AI itself) will be created. AI is important in inequality in three different ways: in destroying jobs (routine based jobs), in creating jobs (data scientists) and in enabling new labor marketplaces to boost the transition from old to new jobs like Catalyte is doing in the US (AI supported new competence building).
Facebook has to deal with this important topic, esp. societally. Why? Research by OECD shows that 20% of unemployment is a result by globalization (China, India, etc.) and the remaining 80% of lost jobs is a result of new technologies, mostly AI / automation.
Overall, it seems Harari is more holistic, global, independent and more alpha than Zuckerberg. Liberal arts is about humans, human nature, wisdom and ethics referring for example to the dialogue around good humans becoming bad or making bad decisions historically. Harari seems to have a better sense of this in this particular dialogue.