Popular Science

Moreover, Conrad managed to contain the complete novella within 65 pages. Aldous Huxley fit “Brave new world” into 249 pages. J. D. Salinger fit “Catcher in the rye” in 198 pages, while George Orwell’s “1984” is approx. 350 pages long (and yes, page count depends on the edition, *obviously*). I seriously worry that in the current environment of “serially-serialised” novels these self-contained masterpieces would struggle to find a publisher. This, plus we’re also seeing true “bigorexia” in literature: novels have been exploding in size in recent years, and probably even more so in the broadly defined sci-fi / fantasy genre. Back when I was a kid, a book over 300 pages was considered long, and “bricks” like Isaac Asimov’s “Foundation” or Frank Herbert’s “Dune” were outliers. Nowadays no one bats an eyelid at 500 or 800 pages.

And this is where Neal Stephenson’s “Fall, or Dodge in hell” comes in. At 883 pages it’s a lot to read through, and I probably wouldn’t have picked it up had it not been referenced by two academics I greatly respect, professors Frank Pasquale and Steve Fuller in their discussion some time ago. I previously read Stephenson’s “Snow crash” as a kid and I was pretty neutral about it: it was an OK cyberpunk novel, but it failed to captivate me; I stuck to Gibson. Still, with such strong recommendations I was more than happy to check “Fall” out.

I will try to keep things relatively spoiler-free. In a nutshell, we have a middle-aged, more-money-than-God tech entrepreneur (the titular Dodge) who dies during a routine medical. His friends and family are executors of his last will, which orders them to have Dodge’s mind digitally copied and uploaded into a digital realm referred to as Bitworld (as opposed to Meatspace, i.e. the real, physical world – btw if you’re thinking “that’s not very subtle”; well, nothing in “Fall” is; subtlety or subtext are most definitely *not* the name of the game here). It takes hundreds of (mostly unnecessary) pages to even get to that point, but, frankly, that part is the book’s only saving grace, because Stephenson hits on something real, which I think usually gets overlooked in the mind uploading discourse: what will it feel to be a disembodied brain in a completely alien, unrelatable, sensory stimuli-deprived environment? This is the one part (only part…) of the novel where Stephenson’s writing is elevated, and we can feel and empathise with the utter chaos and confusion of Dodge’s condition. There was a very, very interesting discussion on a related topic in a recent MIT Technology Review article titled “This is how your brain makes your mind” by psychology professor Lisa Feldman Barrett, which reads “consider what would happen if you didn’t have a body. A brain born in a vat would have no bodily systems to regulate. It would have no bodily sensations to make sense of. It could not construct value or affect. A disembodied brain would therefore not have a mind”. Stephenson’s Dodge is not in the exact predicament Prof. Barrett is describing (his brain wasn’t born in a vat), but given he has no direct memory of his pre-upload experience, it is effectively identical.

One last semi-saving grace is Stephenson’s extrapolated-to-the-extreme a vision of information bubbles and tribes. His America is divided so extremely along the lines of (dis)belief and (mis)information filtering that it is effectively a federation of passively hostile states rather than anything resembling united states. That scenario seems to be literally playing out in front of our eyes.

Unfortunately, Stephenson quickly runs out of intellectual firepower (even though he is most definitely a super-smart guy – after all, he invented the metaverse a quarter of a century before Mark Zuckerberg brought it into vogue) and after a handful of truly original and thought-provoking pages we find ourselves in something between the digital Old Testament and the Medieval, where all the uploaded (“digitally reincarnated”) minds begin to participate in an agrarian feudal society, falling into all the same traps and making all the same mistakes mankind did centuries ago; a sci-fi novel turns fantasy. It’s all very heavy-handed, unfortunately; it feels like Stephenson was paid by the page and not by the book, so he inflated it beyond any reason. If there is any moral or lesson to be taken away from the novel, it escaped me. It feels like the author at one point realised that he cannot take the story much further, or, possibly, just got bored with it and decided to wrap it up.

“Fall” is a paradoxical novel in my eyes: on one hand the meditation on the disembodied, desperately alone a brain is fascinating from the Transhumanist perspective; on the other I honestly cannot recall the last time I read a novel so poorly written. It’s just bad literature, pure and simple – which is particularly upsetting because it is a common offence in sci-fi: bold ideas, bad writing. I have read so many sci-fi books where amazing ideas were poorly written up, and I have a real chip on my shoulder about it, as I suspect that sci-fi literature’s second-class citizen status in the literary world (at least as I have perceived it all my life, perhaps wrongly) might be down to its literary qualities. The one novel that comes to mind as a comparator volume-wise and author’s clout-wise is Neil Gaiman’s “American gods”, and you really need to look no further to see, glaringly, the difference between quality and not-so-quality literature within broadly-defined sci-fi and fantasy genre: Gaiman’s writing is full of wonder with moments of genuine brilliance (Shadow’s experience being tied to a tree) whereas Stephenson’s is heavy, uninspired, and tired.

Against my better judgement, I read the novel through to the end (“if you haven’t read the book back-to-back, then it doesn’t count!” shouts my inner saboteur). Is “Fall” worth the time it takes to go through its 883 pages? No; sadly, it is not. You could read 2 – 3 other, much better books in the time it takes to go through it, and – unlike in that true-life story – there is no grand prize at the end.

What are the lessons to be taken away from 5 months’ worth of wasted evenings? Two, in my view:

Writing a quality novel is tough, but coming up with a quality, non-WTF ending is tougher; that is where so many fail (including Stephenson – spectacularly);
If a book isn’t working for you, just put it down. Sure, it may have a come to Jesus revelatory ending, but… how likely is that? Bad novels are usually followed by even worse endings.

_______________________________________________________

(1)  Another case in point: Thomas Harris’ Hannibal Lecter series of novels – it’s mediocre literature at best, but it allowed a number of very talented artists to develop fascinating, charismatic characters and compelling stories, both in cinema and on TV.

(2) In my personal view, both novels are on the edge of fantasy and slipstream, but I appreciate that not everyone will agree with this one.

“Reflecting forward” occupies the entirety of Moco’s basement (and frankly, is quite easy to miss), which is a brilliant idea, because it blocks off all external stimuli and allows the visitors to fully immerse themselves in Irma’s work. Funny thing: main floors are filled with works by some of modern art’s most accomplished heavyweights (Banksy, Warhol, Basquiat, Keith Haring – even a small installation by Yayoi Kusama), and yet lesser-known Irma completely blows them out of the water.

I don’t know how from a seasoned art critic’s perspective “Reflecting forward” compares to Van Goghs or Rembrandts just a stone’s throw away (or even Warhols and Banksys upstairs at the Moco) in terms of artistic highbrow-ness. I’m sure it’d be an interesting, though ultimately very subjective discussion. Having visited both the Rijksmuseum and the Moco, I was captivated by “Reflecting forward” more than by classical Dutch masterpieces (with the exception of “Night watch”, which is… just mind-blowing). 

With “Reflecting forward” Irma is promoting a new art movement, called connectivism, defined as “a theoretical framework for understanding and learning in a digital age. It emphasizes how internet technologies such as web browsers, search engines, and social media contribute to new ways of learning and how we connect with each other. We take this definition offline and into our physical world. through compassion and empathy, we build a shared understanding, in our collective choice to experience art”¹.

An immersive experience is only immersive as a full package. Ambience and sound play a critical (even if at times subliminal) play a critical part in the experience. Studio Irma commissioned a bespoke soundscape for “Reflecting forward”, which takes the experience to the whole new level of dreamy-ness. The music is really quite hypnotic. The artist stated her intention to release it on Spotify, but as of the time of writing, it is not there yet.

I believe there is also an augmented reality (AR) app allowing us to take the “Reflecting forward” outside, developed with socially-distanced, pandemic-era art experience in mind, but I couldn’t find the mention of it on Moco’s website, and I haven’t tried it.

Overall, for the time I stayed in Moco’s basement, “Reflecting forward” has transported me to beautiful, peaceful, and hypnotic places, giving me an “out of time” experience. Moco deserves huge kudos for giving the project the space it needed and allowing the artist to fully realise her vision. Irma de Vries’ talent and imagination shine in “Reflecting forward”, and I hope to experience her work in the future.

______________________________________

¹ http://studioirma.com/biography.html 

Then came the Web, followed by smartphones.

Nowadays, with smartphones and Google, finding information online is so easy and immediate that many of us have forgotten how difficult it was a mere 20-something years ago. Moreover, more and more people have simply never experienced this kind of existence at all.

We get more content pushed our way than we could possibly consume: Chrome Discover recommendations; every single article in your social media feeds; all the newsletters you receive and promise yourself to start reading one day (spoiler alert: you won’t; same goes for that copy of Karl Marx’s “Capital” on your bookshelf); all the articles published daily by your go-to news source; etc. Most people read whatever they choose to read, close the tab, and move on. Some remember what they’ve read very well, some vaguely, and over time we forget most of it (OK, so it probably is retained on some level [like these very interesting articles claim ¹,²], gradually building up our selves much like tiny corals build up a huge coral reef, but hardly anyone can recall titles and topics of the articles they read days, let alone years prior).

A handful of the articles we read will resonate with us really strongly over time. Some of us (not very many, I’m assuming) will Evernote or OneNote them as their personal, local-copy notes; some will bookmark them; and some will just assume they will be able to find these articles online with ease. My question to you is this: how many times have you tried looking for that article you wanted to re-read only to find yourselves losing it an hour or two later, unsure as to whether you’ve ever read that article in the first place or whether that has been just a figment of your imagination? (if your answer is “never”, you can probably skip the rest of this post). It happened to me so many, many times that I started worrying that perhaps something *is* wrong with my head.

And Web pages are not even the whole story. Social media content is even more “locked” within the respective apps. Most (if not all) of them can technically be accessed via the Web and thus archived, but this rarely works in practice. It works for LinkedIn, because LI was developed as browser-first application, and it’s generally built around posts and articles, which lend themselves to browser viewing and saving. Facebook was technically developed as browser-first, but good luck saving or clipping anything. FB is still better than Instagram, because with FB we can still save some of the images, whereas with Instagram that option is practically a non-starter. Insta wants us to save our favourites within the app, thus keeping it fully under its (i.e. Meta’s) control. That means that favourited images can at any one time be removed by the posters, or by Instagram admins, without us ever knowing. I don’t know how things work with Snap, TikTok, and other social media apps, because I don’t use them, but I suspect that the general principle is similar: content isn’t easily “saveable” outside the app.

Then there are ebooks, which are never fully offline the way paper books are. The Atlantic article ³ highlights this with a hilarious example of the word “kindle” being replaced with “Nook” in a Nook e-reader edition of… War and Peace.

Then came Iza Kaminska’s 2017 FT article “The digital rot that threatens our collective memory” and I realised that maybe nothing’s wrong with my head, and that maybe it’s not just me. While Iza’s article triggered me, its follow-up in The Atlantic nearly four years later (“The Internet is rotting”) super-triggered me. The part “We found that 50 percent of the links embedded in Court opinions since 1996, when the first hyperlink was used, no longer worked. And 75 percent of the links in the Harvard Law Review no longer worked” felt like a long-overdue vindication of my memory (and sanity). It really wasn’t me – it was the Internet.

It really boggles the mind: in just 2 – 3 decades the Web has replaced libraries and physical archives as humankind’s primary source of reference information (and arguably fiction as well, though paper books are still holding reasonably strong for now). The Internet is comprised of websites, and websites are comprised of posts and articles – all of which have a unique reference in the form of the URL. I can’t talk for other people, but I have always implicitly assumed that information lived pretty much indefinitely online. On the whole it may be the case (the reef so to speak), but that does not hold for specific pages (individual corals) anywhere near as much as I had assumed. There is, of course, the Internet Archive’s Wayback Machine, a brilliant concept and a somewhat unsung hero of the Web – but WM is far from solid (example: I found a dead link [ http://www.psychiatry.cam.ac.uk/blog/2013/07/18/bad-moves-how-decision-making-goes-wrong-and-the-ethics-of-smart-drugs/ ] in one of my blog posts [“Royal Institution: How to Change Your Mind”]. The missing resource was posted on one of the websites within the domain of Cambridge University, which indicates it is high-quality, valuable, meaningful content – and yet the Wayback Machine has not archived it. This is not a criticism of WM nor its parent the Internet Archive – what they do deserves the highest praise; it’s more about recognising the challenges and limitations it’s facing).

So, with my sanity vindicated, but my OCD and FOMO being as voracious as ever – where do we go from here?

There is the local copy / offline option with Evernote, OneNote, and similar Web clippers. I have been using Evernote for years now (this is not an endorsement), and, frankly, it’s far from perfect, particularly on a mobile (I said this was not an endorsement…) – but everything else I have come across has been even worse; and, frankly, there are surprisingly few players in that niche. Still, Evernote is the best I could find for all-in-one article clipping and saving – it’s either this or “save article as”, which is a bit too… 90’s. Then there is the old-fashioned direct e-mail newsletter approach, which, as proven by Substack, can still be very popular. I’m old enough to remember the life pre-cloud (or even pre-Web, albeit very faintly, and only because the Web arrived in Poland years after it arrived in some of the more developed Western countries) – C:\ drive, 1.44Mb floppy disks, and all that – and that, excuse the cheap pun, clouds my judgement: I embrace the cloud and I love the cloud, but I want to have a local, physical copy of my news archive. However, as Drew Justin rightly points out in his Wired article “As with any public good, the solution to this problem [the deterioration of the infrastructure for publicly available knowledge] should not be a multitude of private data silos, only searchable by their individual owners, but an archive that is organized coherently so that anyone can reliably find what they need”; me updating my personal archive in Evernote (at the cost of about 2hrs of admin per week) is a crutch, an external memory device for my brain. It does nothing for the global, pervasive link rot. Plus, Evernotes of this world don’t work with social media apps very well. There are dedicated content clippers / extractors for different social media apps, but they’re usually a bit cumbersome, and don’t really “liberate” the “locked” content in any meaningful way.

I see two ways of approaching this challenge:

• Beefing up the Internet Archive, the International Internet Preservation Consortium, and similar institutions to enable duplication and storage of as wide a swath of Internet as possible, as frequently as possible. That would require massive financial outlays and overcoming some regulatory challenges (non-indexing requests, the right to be forgotten, GDPR and GDPR-like regulations worldwide). Content locked in-app would likely pose a legal challenge to duplicate and store if the app sees itself as the owner of all the content (even if it was generated by the users).
• Accepting the link rot as inevitable and just letting go. It may sound counterintuitive, but humankind has lost countless historical records and works of art from the sixty or so centuries of the pre-Internet era and somehow managed to carry on anyway. I’m guessing that our ancestors may have seen this as inevitable – so perhaps we should too? 

I wonder… is this ultimately about our relationship with our past? Perhaps trying to preserve something indefinitely is unnatural, and we should focus on the current and the new rather than obsess over archiving the old? This certainly doesn’t agree with me, but perhaps I’m the one who’s wrong here…

________________________________________________

¹https://www.wired.co.uk/article/forget-idea-we-forget-anything 

² https://www.wired.com/2009/09/forgottenmemories/ 

³ https://www.theatlantic.com/technology/archive/2021/06/the-internet-is-a-collective-hallucination/619320/ 

Prototype of CO2-capturing machine, Science Museum, London 2022

Geoengineering has all the attributes needed to make it into the zeitgeist: a catchy, self-explanatory name; explicit hi-tech connotations; and the potential to become the silver bullet (at least conceptually) that almost magically balances the equations of still-rising emissions with the desperate need to keep the temperature rise as close to 1.5 Centigrade as possible.

The Cambridge Zero event was a great introduction to the topic and its many nuances. Firstly, there are (at least) two types of geoengineering:

• Solar (increasing the reflectivity of the Earth to reflect more and absorb less of the Sun’s heat in order to reduce the planet’s temperature);
• Carbon (removing the already-emitted excess CO2 from the atmosphere).

The broad premise of solar geoengineering is to scatter sunshine in the stratosphere (most likely by dispersing particles of highly reflective compounds or materials). While some compare it to a result of a volcano explosion, the speaker’s suggestion was to compare it to a thin layer of global smog. The conceptual premise of solar geoengineering is quite easy to grasp (which is not the same as saying that solar geoengineering itself is in any way easy – as a matter of fact, it is extremely complex and, in the parlance of every academic paper ever, “further research is needed”). The moral and political considerations may be almost as complex as the process itself. There is a huge moral hazard that fossil fuel industry (and similar vested interests, economic and political) might perform a “superfreak pivot”, going from overt or covert climate change denial to acknowledging it and pointing to solar geoengineering as the only way to fix it. Consequently, these entities and individuals would no longer need to deny the climate is changing (which is becoming an increasingly difficult position to defend these days), while they could still push for business as usual (as in: *their* business as usual) and delay decarbonisation.

The quote from the book Has It Come to This: The Promises and Perils of Geoengineering on the Brink puts is brilliantly: “Subjectively and objectively, geoengineering is an extreme expression of a – perhaps *the* – paradox of capitalist modernity. The structures set up by people are perceived as immune to tinkering, but there is hardly any limit to how natural systems can be manipulated. The natural becomes plastic and contingent; the social becomes set in stone.”

The concept is carbon geoengineering is also very straightforward (conceptually): to remove excess carbon from the atmosphere. There are two rationales for carbon geoengineering:

• To offset carbon emissions that – for economic or technological reasons – cannot be eliminated (emissions from planes are one example that comes to mind). Those residual emissions will be balanced out by negative emissions resulting from geoengineering;
• To compensate for historical emissions.

The IPCC special report on 1.5°C states unambiguously that limiting global warming to 1.5°C needs to involve large-scale carbon removal from the atmosphere (“large-scale” being defined as 100 – 1,000 gigatons over the course of the 21st century). This, in my view, fundamentally differentiates carbon geoengineering from solar geoengineering in terms of politics and policy: the latter is more conceptual, a “nice to have if it works” lofty concept; the former is enshrined in climate policy plan from the leading scientific authority on climate change. This means that carbon capture is not a “nice to have”, it’s critical.

Carbon geoengineering is a goal that can be achieved using a variety of means: from natural (planting more trees, planting crops with roots trapping more carbon) to technological (carbon capture and storage). The problem is that it largely remains in the research phase, and is nowhere near deployment at scale (which, on carbon capture and storage side is akin to having a parallel energy infrastructure in reverse).

There is an elephant in the room that shows the limitations of geoengineering: the (im)permanence of the results. The effects of solar geoengineering are temporary, while carbon capture has its limits: natural (available land; be it for carbon-trapping vegetation or carbon capture plants), technological (capture and processing plants), and storage. Geoengineering could hopefully stave off the worst-case climate change scenarios, slow down the rate at which the planet is warming, and / or buy mankind a little bit of much-needed time to decarbonize the economy – but it’s not going to be a magic bullet.

Stanisław Lem in 1966

The thing about Communism (and that’s speaking from limited and still-learning perspective) is that there was in fact much, much more to it than many people realise. We’re talking decades (how many decades exactly depends on individual country) and hundreds of millions of people, so that’s obviously a significant part of the history of the 20th century. The Iron Curtain held so tight that for many years the West was either missing out altogether or had disproportionately low exposure to culture, art, or science of the Eastern Bloc (basically everything that was not related to the Cold War). As the West was largely about competition (including competition for attention) there was limited demand for Communist exports because there wasn’t much of a void to fill. That doesn’t mean that there weren’t exciting ideas, philosophies, works of art, or technological inventions being created in the East.

The GAIN event focused on the fascinating intersection of philosophy, literature, and technology. It just so happens that one of the (world’s) most prolific Cold War-era thinkers on the topic of the future of technology and mankind in general was Polish. I’m referring to the late, great, there-will-never-be-anyone-like-him Stanislaw Lem (who deserved the Nobel Prize in Literature like there was no tomorrow – one could even say more than some of the Polish recipients thereof). Lem was a great many things: he was a prolific writer whose works span a very wide spectrum of sci-fi (almost always with a deep philosophical or existential layer), satire disguised as sci-fi, and lastly philosophy and technology proper (it was in one of his essays in the late 1990’s or early 2000’s I have first read of the concept of the brain-computer interface (BCI); I don’t know to what extent BCI was Lem’s original idea, but he was certainly one of its pioneers and early advocates). He continued writing until his death in 2006.

One of Lem’s foremost accomplishments is definitely 1964’s Summa Technologiae, a reference to Thomas Acquinas’ Summa Theologiae dating nearly seven centuries prior (1268 – 1273). Summa discusses technology’s ability to change the course of human civilisation (as well as the civilisation itself) through cybernetics, evolution (genetic engineering), and space travel. Summa was the sole topic of one of the GAIN event’s presentations, delivered by Bogna Konior, an Assistant Arts Professor at the Interactive Media Arts department of NYU Shanghai. Konior took Lem’s masterpiece out of its philosophical and technological “container” and looked at it from a wider perspective of the Polish social and political system Lem lived in – a system that was highly suspicious, discouraging (if not openly hostile) to new ways of thinking. She finds Lem pushing back against the political status quo.

While Bogna Konior discussed one of the masterpieces of a venerated sci-fi giant, the next speaker, Jędrzej Niklas, presented what may have sounded like sci-fi, but was in fact very real (or at least planned to happen for real). Niklas told the story of Poland’s National Information System (Krajowy System Informatyczny, KSI) and a (brief) eruption of “technoenthusiasm” in early 1970’s Poland. In a presentation that sounded at times more like alternative history than actual one Niklas reminded us of some of the visionary ideas developed in Poland around late 1960’s / early 1970’s. KSI was meant to be a lot of things:

• a central-control system of the economy and manufacturing (pls note that at that time vast majority of Polish enterprises were state-owned);
• a system of public administration (population register, state budgeting / taxation, natural resources management, academic information index and search engine);
• academic mainframe network;
• “Info-highway” – a broad data network for enterprises and individuals linking all major and mid-size cities.

If some or all of the above sound familiar, it’s because they all became everyday uses cases of the Internet. [sidebar: while we don’t / can’t / won’t know for sure, there have been some allegations that Polish ideas from the 1970’s were duly noted in the West; whether they became an inspiration to what ultimately became the Internet we will never know].

While KSI ultimately turned out to be too ambitious and intellectually threatening for the ruling Communist Party, it has not been a purely academic exercise. The population register part of KSI became the PESEL system (an equivalent of the US Social Security Number or British National Insurance Number), which is still in use today, while all the enterprises are indexed by the REGON system.

And just like that, the GAIN / LCFI event made us all aware how many ideas which have materialised (or are likely to materialise in the foreseeable future) may not have originated exclusively in the Western domain. I’m Polish, so my interest and focus are understandably on Poland, but I’m sure the same can be said by people in other, non-Western, parts of the world. While the GAIN / LCFI events have not been recorded in their entirety (which is a real shame), they will form a part of the forthcoming book “Imagining AI: how the world sees intelligent machines” (Oxford University Press). It’s definitely one to add to cart if you ask me.

____________________________________

1. I don’t think that any single work of Lem’s can be singled out as his ultimate masterpiece. His best-known work internationally is arguably Solaris, which had two cinematic adaptations (by Tarkovsky and by Soderbergh), which is equal parts sci-fi and philosophy. Summa Technologiae is probably his most venerated work in the technology circles, possible in the philosophy circles as well. The Star Diaries are likely his ultimate satirical accomplishment. Eden and Return from the Stars are regarded as his finest sci-fi works.

Only exceptionally, exceedingly rarely does an academic article make its way into the Zeitgeist (off the top of my head I can think of one); countless articles fade into obscurity having been read only by the authors, their loved ones, journal reviewers, editors, and no one else ever (am I speaking from experience? Most likely). I sometimes wonder how much truly world-changing research is out there, collecting (digital) dust, forgotten by (nearly) everyone.

I have no ambition of turning my blog into a recommender RSS feed, but every now and then, when I come across something truly valuable, I’d like to share it. Such was the case with “theory of mind in AI” article (disclosure: Prof. Cuzzolin is my PhD co-supervisor, while Prof. Sahakian is my PhD supervisor).

Published in Psychological Medicine, the article encourages approaching Artificial Intelligence (AI), and more specifically Reinforcement Learning (RL) from the perspective of hot cognition. Hot and cold cognition are concepts somewhat similar to very well-known concepts of thinking fast and slow, popularised by the Nobel Prize winner Daniel Kahneman in his 2014 bestseller. While thinking fast focuses on heuristics, biases, and mental shortcuts (vs. fully analytical thinking slow), hot cognition is describing thinking that is influenced by emotions. Arguably, a great deal of human cognition is hot rather than cold. Theory of Mind (ToM) is a major component of social cognition which allows us to infer mental states of other people.

By contrast, AI development to date has been overwhelmingly focused on purely analytical inferences based on vast amounts of training data. The inferences are not the problem per se – in fact, they do have a very important place in AI. The problem is what the authors refer to as “naïve pattern recognition incapable of producing accurate predictions of complex and spontaneous human behaviours”, i.e. the “cold” way these inferences were made. The arguments for incorporating hot cognition, and specifically ToM are entirely pragmatic and include improved safety of autonomous vehicles, more and better applications in healthcare (particularly in psychiatry), and likely a substantial improvement in many AI systems dealing directly with humans or operating in human environments; ultimately leading to AI that could be more ethical and more trustworthy (potentially also more explainable).

Whilst the argument in favour of incorporating ToM in AI makes perfect sense, the authors are very realistic in noting limited research done in this field to date. Instead of getting discouraged by it, they put forth a couple of broad, tangible, and actionable recommendations on how one could get to Machine Theory of Mind whilst harnessing existing RL approaches. The authors are also very realistic in challenges the development of Machine ToM will likely face, such as empirical validation (which will require mental state annotations to learning data to compare them with the mental states inferred by AI) or performance measurement. 

I was very fortunate to attend one of co-author’s (Bogdan-Ionut Cirstea) presentation in 2020, which was a heavily expanded companion piece to the article. There was a brilliant PowerPoint presentation which dove into the AI considerations raised in the article in much greater detail. I don’t know whether Dr. Cirstea is at liberty to share it, but for those interested in further reading it would probably be worthwhile to ask.

You can find the complete article here; it is freely available under Open Access license.