I actually have been following the Polish AI startup scene for some time, and I actually have decided to structure the knowledge and create a landscape of Polish AI startups —

This landscape goals to discover and categorize progressive AI startups from Poland. I also hope the landscape will make it easier for featured corporations to access capital, clients, and talent.

There was plenty of hype around AI recently, and almost every startup claims to be an AI startup. Keeping that in mind, I did my best to pick only startups which supply solutions by which AI is an inherent element. In other words, if you happen to can remove the AI component from the answer provided by the corporate and the answer’s value from the client’s perspective doesn’t decrease dramatically, the corporate is excluded from the list.

Over the previous couple of years, I actually have observed a considerable increase in the attention and maturity of AI-based startups in Poland. For a lot of corporations, AI is not any longer only a buzzword used to draw clients and investors but a critical differentiator of value added as a substitute. This fact, combined with the extraordinary level of technical skills and talent, makes our region a really interesting place to begin, grow and scale AI-based organizations.

Kamil’s words are confirmed by the map you may see below. There are already 100 AI startups in Poland and I expect this number to only increase in the following few years.

the landscape, we see that the healthcare category accommodates probably the most AI startups. This is just not surprising since additionally it is the category that received probably the most VC funding in Poland — . One other significant category is customer support, chat, and voice. The 2 categories contain a number of the most successful startups from the list: Infermedica (120m PLN series B in 2021) and Tido (109m PLN series B in 2022).

is certainly the most well liked AI topic nowadays, yet the myriad of recent applications it unlocks justifies the hype. I expect to see many latest startups emerge because of this technology. might be probably the most obvious category that might be affected by progress in generative AI because of the conversation-like nature of interactions with customers inherent to tools like Chat GPT. I believe we may also see many latest startups working on that can enable cheaper and faster content generation. This is just not only limited to text since .

may also likely flourish in the approaching years. Now we have already seen applications that tease the potential of this technology e.g. Codex or Copilot . Nonetheless, I believe we’re still yet to see plenty of latest tools that make creating, testing, or debugging software easier and faster.

Also, I see immense potential in that enable less technical users to create AI applications. This might be fueled by the but a really

One other category I expect to grow considerably is AI in healthcare. That is already the largest category within the Polish ecosystem, but . There’s an explosion of health data because of electronic medical records, insurance claims, the myriad of health-tracking apps, and wearables. All of them provide data sources for training AI models (based on some estimates, ). Also, digital AI healthcare solutions are getting more accessible because of virtual and mobile care. Because of this, patients are getting more accustomed to tech-enabled healthcare, which can increasingly be AI-powered.

One major trend that can profit all AI applications is the . As an example, based on research by Stanford, the associated fee of coaching a picture classification system has decreased by 2/3 since 2018, with an excellent larger decrease in model training time.

Cost decrease is especially possible because of akin to Google’s Tensor Processing Units (TPUs) or Nvidia’s chips e.g. A100. For AI corporations, cloud computing costs are frequently a serious COGS (cost of products sold) component. When computing costs decrease, they make latest applications economically viable, thus .

(estimated training cost of GPT-3 is as much as 12m USD). This trend can offset the effect of decreasing unit computing costs, yet this concerns mostly state-of-the-art models published by researchers reasonably than already commercialized systems.

(i.e. inability to know how an algorithm makes decisions) of many AI systems is certainly one of the numerous barriers to AI adoption. Users often require a prediction from the algorithm and a proof of how the prediction was made. Offering this explanation is incredibly hard for algorithms akin to neural networks, and the dearth of the reason may prevent some users from adopting AI solutions. Increasing numbers of models’ parameters, only amplify this problem.

One other challenge is present within the training data set. As an example, one generative AI algorithm asked to generate images of “a poor person” and “thug” generated predominately faces with dark skin. If adopted on a large scale, biased algorithms will amplify stereotypes and biases already present in our society.

data can also be a priority. Despite a growing amount of knowledge generated globally, data stored by businesses is incessantly of poor quality, making the implementation of AI systems much tougher.

Despite these challenges, I’m convinced that Polish and global AI ecosystems have one of the best years ahead. I cannot wait to see latest startups that can emerge!

In case you know any company that needs to be added to the list, fill on this form and let me know within the comments!