Wed 29-Apr-2020
Countries most synonymous with “AI powerhouses” are without a doubt the US and China. Both have economies of scale, resources, and strategic (not just business) interests in being at the forefront of AI. EU as a whole would probably come third, although there is always a degree of subjectivity in these rankingsi. The UK would probably come next (owing to Demis Hassabis and DeepMind, as well as thriving scientific and academic communities). In any case, it’s rather unlikely that Poland would be listed in the top tier. Poland is known for being an ideal place to set up corporate back- or (less frequently) middle office functions: much cheaper than Western Europe, with huge pool of well-educated talent, in the same time zone as the rest of the EU. A great alternative (or complement) to setting up a campus in India, but not exactly a major player in AI research and entrepreneurship. Plus, Poland and its young democracy (dating back to 1989) are currently going through a bit of a social, identity, and political rough patch. Not usually a catalyst or enabler of cutting-edge technology.
And despite all that (and despite being a mid-sized country at best… 38 million people; and despite being global number #70 globally in GDP per capita in 2018 out of 239 countries and territoriesii) for some mysterious reason Poland still made it to #15 globally in AI (using total number of companies as a metric) according to China Academy of Information and Communications Technology (CAICT) Data Research Centre Global Artificial Intelligence Industry Data Reportiii as kindly translatediv by my fellow academic Jeffrey Ding from Oxford University (whose ChinAI newsletter is brilliant and I encourage everyone to subscribe and read). I found this news so unexpected that it was the inspiration behind the entire post below.
The recent (2019) Map of the Polish AI from Digital Poland Foundation reveals a vibrant, entrepreneurial ecosystem with a number of interesting characteristics. The official Polish AI Development Policy 2019 – 2027 released around the same time by a multidisciplinary team working across a number of government ministries paints a picture of impressive ambitions, though experts have questioned their realism.
The Polish AI scene is very young (50% of the 160 organisations polled introduced AI-based services in 2017 or 2018, the most recent years in the survey). Warsaw (unsurprisingly) steals the top spot, with 85% of all companies being located in one of the 6 major metropolitan areas. The companies tend to be small: only 22% have more than 50 people; 59% have 20 or fewer. Let’s not conflate company headcount with AI teams proper – over 50% of the companies surveyed have AI teams of 5 employees or fewer. Shortage of talent is a truly global theme in AI (which I personally don’t fully agree with – companies with resources to offer competitive packages [sometimes affectionately referred to as “basketball player salaries”] have no shortage of candidates; whether this level of pay is justifiable [the very short-lived bonanza for iOS app developers circa 2008 comes to mind] and fair to the smaller players is a different matter). The additional challenge in Poland is that Polish salaries simply cannot compete with what is on offer within 3 hours’ flight – many talented computer scientists are naturally tempted to move to places like Berlin, Paris, London or other major European AI hubs where there are more opportunities, more developed AI ecosystems, and much, much better money to be made.
What stands out is the ultra-close connection between business and academic communities. While the same is the case in most countries seriously developing AI, some of them are home to global tech corporates whose financial resources and thus R&D capabilities give them the luxury to develop on their own, at par (if not ahead) of leading research institutions. These corporates’ resources also enable them to poach world-class talent (e.g. Google hiring John Martinis to lead their quantum computer efforts [who has since left…], Facebook appointing Yann LeCun as head of AI research, or Google Cloud poaching [albeit briefly] Fei-Fei Li as their Chief Scientist of AI/ML). In Poland this will not apply – it does not have any large (or even mid-size) home-grown innovative tech firms. The ultra-close connection between business and academia is a logical consequence of these factors – plus in a 38-million-strong country with relatively few major cities serving as business and academic hubs, the entire ecosystem simply can’t be very populous.
The start-up scene might in part be constrained by limited amount of available funds (anecdotally the angel investor / VC scene in Poland is very modest). However, the Digital Poland report states:
Categorically, as experts point out, the main barrier to the development of the Polish AI sector is not the absence of funding or expertise but rather a very limited demand for solutions based on AI.
My personal contacts echo this conclusion – they are not that worried about funding. Anecdotally, there is a huge pool of state grants (NCiBR) with limited competition for those (although post-COVID-19 they may all but evaporate).
Multiple experts cited by Digital Poland all list domestic demand as the primary concern. According to the survey, potential local clients simply do not understand the technology well enough to realise how it can benefit them (41% of responses in a multiple choice questionnaire – single highest cause; [client] staff not understanding AI had its own mention at 23% and [managers] not understanding AI came at 22%).
The AI market in Poland is focused on more commercial products (Big Data analytics, sales, analytics) rather than cutting-edge innovative research. It is understandable – in an ecosystem of limited size with very limited local demand the start-ups’ decision to develop more established, monetisable, applications which can be sold to a broad pool of global clients is a reasonable business strategy.
One side-conclusion I found really interesting is that there’s quite a vibrant conference and meetup scene given how nascent and “unsolidified” the AI ecosystem is.
The Polish AI Policy document is an interesting complement to the Digital Poland report. While the latter is a thoroughly researched snapshot of the Polish AI market right here right now (2019 to be exact), the policy document is more of a mission statement – a mission of impressive ambitions. I always support bold, ambitious, and audacious thinking – but experience has taught me to curb my enthusiasm as far as Polish policy-making is concerned. Grand visions of the 2019 – 2027 come with not even a draft of a roadmap. The document is also, unfortunately, quite pompous and vacuous at times.
The report is rightly concerned about impact on jobs, concluding that the expectation is that more jobs will be created than lost, and concluding that some of this surplus should benefit Poland. One characteristic of Polish economy is that it (still) has a substantial number of state-owned enterprises in key industries (banking, petrochemicals, insurance, mining and metallurgy, civil aviation, defence), which are among the largest in their industries on a national scale. Those companies have the size and the valid business cases for AI, yet they don’t seem ready (from education and risk-appetite perspectives) to commit to AI. State-level policy could provide the nudge (if not outright push) towards AI and emerging technologies, yet, unfortunately, that is not happening.
The report rightly acknowledges the skills gap, as well as some issues on the education side (dwindling PhD rates, relatively low (still!) level of interest in AI among Polish students as measured by thesis subject choices). The quality of Polish universities merits its own article (merits its own research, in fact). On one hand, the anecdotal and first-hand experiences lead me to believe that Polish computer scientists are absolutely top-notch, on the other, the university rankings are… unforgiving (there are literally two Polish universities on QS Global 500 list 2020, at positions #338 and #349v).
Last but not least, a couple of Polish AI companies I like (selection entirely subjective):
- Sigmoidal – AI business/management consultancy.
- Edward.ai – AI-aided sales and customer relationship management (CRM) solutions.
- Fingerprints.digital – behavioural biometrics solutions.
Disclaimer: I have no affiliations with any of the abovementioned companies.
[i] Are we looking at corporate research spending? Government funding/grants for academia? Absolute amounts or % of GDP? How reliable are the figures and how consistent are they between different states? etc. etc.
[ii] Source: World Bank (https://data.worldbank.org/indicator/NY.GDP.PCAP.CD?most_recent_value_desc=true)
[iii] You can read the original Chinese version here: http://www.caict.ac.cn/kxyj/qwfb/qwsj/201905/P020190523542892859794.pdf
[iv] Jeff’s English translation can be found here: https://docs.google.com/document/d/15WwZJayzXDvDo_g-Rp-F0cnFKJj8OwrqP00bGFFB5_E/edit#heading=h.fe4k2n4df0ef
[v] https://www.topuniversities.com/university-rankings/world-university-rankings/2020
