Implementing AI-based automation – practical challenges

There is an anecdote circulating in the IT industry and among lawyers dealing with artificial intelligence. A contractor talking about the advantages of his solution focuses on how smart it is and repeatedly refers to the term ‘artificial intelligence’. After the presentation, the contracting authority asks if the contractor is following the development of the legal ecosystem that is supposed to regulate artificial intelligence and how he ensures that his AI solution is ready for new regulations. It is then answered by a slightly dismayed contractor that this is actually… an advanced algorithm, not AI at all.

This anecdote captures the first two significant practical challenges faced by those involved in implementing automation.

AI or not AI?

The first is related to which solutions are actually dealing with Artificial Intelligence (i.e. software that is actually characterised by the ability to learn and solve problems on its own – without the use of pre-programmed algorithms). This question will often be answered by engineers and programmers, provided that the contractor realistically allows them to get acquainted with the code and architecture of the system.

Unregulated regulation

The second challenge concerns the regulatory ecosystem. Currently in Poland and in the EU there are no universally binding rules (laws, regulations or EU directives) that would unambiguously determine what is allowed and what is not allowed when creating and using artificial intelligence. There are also no binding guidelines on the characteristics and requirements that technical solutions based on artificial intelligence must meet. There are white papers[1], communications[2], strategies[3], assumptions and draft legislation[4]. None of these are (at the time of writing) binding laws, the solutions they propose often differ, and lobbyists and interests of different groups often clash. There is no guarantee when and exactly in what form binding regulations will come into force. Of course, there are regulations that can be used auxiliary (general principles of civil law, GDPR, etc.), but this means that when designing and implementing AI solutions, one has to be ready to be flexible and to react quickly and efficiently to developments and legislative changes.

„Without data, there is no AI”

However, the two challenges described so far do not end there. Another is related to the fact that, as rightly pointed out in the 2020 EU White Paper on AI, “Without data, there is no AI.” So, one of the key aspects for the effectiveness of artificial intelligence is the data on which the algorithms “train themselves” and from which they draw conclusions. Consequently, the data that is fed to the algorithms is as important as the program itself. The latter may work fully correctly, while due to incomplete, manipulated or biased data the effect of the complete solution will be absolutely unacceptable.

A good example is the experiment with the Twitter chatbot Tay in 2016. The developers at Microsoft had to turn it off after several hours because the AI started responding to the TT users accosting it in a way it had learned from them – far from expected and acceptable. This case shows that the question of whose and what quality of data we will teach the algorithms can be a decisive factor when choosing a contractor or negotiating a contract.

Black box formula

For very sophisticated and advanced solutions based on artificial intelligence, it is often offered in a “black box” formula. In practice, this means that we get conclusions and results, but we cannot trace the process or answer the question why we got this result and not another. Sometimes the reason lies in the high complexity, and usually in the desire to protect the contractor’s know-how and code. This makes it difficult, and sometimes impossible, to be accountable and to ensure that the nature of AI performance is non-discriminatory.

In conclusion, when implementing solutions based on artificial intelligence in a company or institution, one has to set oneself up for a fascinating journey into often still unknown regions. Therefore, it is worth preparing well for it, taking care of data quality and nuances in the contract and actively following the changing legal ecosystem.

[1] EC, White Paper on Artificial Intelligence. A European approach to excellence and trust, COM (2020) 65 final, 19.02.2020.

[2] Communication from the Commission to the European Parliament, the European Council, the European Economic and Social Committee and the Committee of the Regions: Artificial Intelligence for Europe, COM (2018) 237 final, 25.04.2018; Communication from the Commission to the European Parliament, the European Council, the Council, the European Economic and Social Committee and the Committee of the Regions: A coordinated plan on artificial intelligence, COM (2018) 795 final, 07.12.2018; Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions: Promoting a European approach to Artificial Intelligence, COM (2021) 205 final, 21.04.2021.

[3] Resolution No. 196 of the Council of Ministers of 28 December 2020 on establishing a “Policy for the development of artificial intelligence in Poland from 2020”.

[4] Proposal for a Regulation of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) and amending certain legislative acts of the Union, COM(2021) 206 final, 21.04.2021.

The article was published in the monthly magazine Automatyka, issue 1-2/2022. – https://automatykaonline.pl/Automatyka/Roczniki/2022/1-2-2022

The author – Maciej Kubiak – specializes in copyright, media, film and press law, IT and new technologies, advertising and industrial property and protection of personal rights. He represents clients in court disputes related to the protection of intellectual property, reputation and in commercial matters. He has completed dozens of legal services projects for film and television productions, including high-budget international co-productions. He has advised many producers, music and book publishers, concert organisers, advertising and public relations agencies, IT companies and media houses. He consults on projects related to innovation, implementation of automation and robotics, personal data, new technologies and commercialisation of scientific research. He supports clients in matters related to fighting unfair competition. He has assisted in cross-border licensing projects and the entry of new brands on the Polish market, including new TV channels.

You can find more articles on similar topics on our blog.