Ladies in AI: Emilia Gómez on the EU began her AI profession with music

To provide AI-focused girls lecturers and others their well-deserved — and overdue — time within the highlight, TechCrunch is launching a series of interviews specializing in outstanding girls who’ve contributed to the AI revolution. We’ll publish items all year long because the AI increase continues, highlighting key work that always goes unrecognized. Learn extra profiles here.

Emilia Gómez is a principal investigator on the European Fee’s Joint Analysis Centre and scientific coordinator of AI Watch, the EC initiative to observe the developments, uptake and impression of AI in Europe. Her crew contributes with scientific and technical data to EC AI insurance policies, together with the not too long ago proposed AI Act.

Gómez’s analysis is grounded within the computational music subject, the place she contributes to the understanding of the way in which people describe music and the strategies by which it’s modeled digitally. Ranging from the music area, Gómez investigates the impression of AI on human conduct — particularly the results on jobs, selections and youngster cognitive and socioemotional growth.

Q&A

Briefly, how did you get your begin in AI? What attracted you to the sector?

I began my analysis in AI, particularly in machine studying, as a developer of algorithms for the automated description of music audio indicators by way of melody, tonality, similarity, fashion or emotion, that are exploited in several purposes from music platforms to schooling. I began to analysis how one can design novel machine studying approaches coping with completely different computational duties within the music subject, and on the relevance of the info pipeline together with knowledge set creation and annotation. What I appreciated on the time from machine studying was its modelling capabilities and the shift from knowledge-driven to data-driven algorithm design — e.g. as an alternative of designing descriptors primarily based on our data of acoustics and music, we have been now utilizing our know-how to design knowledge units, architectures and coaching and analysis procedures.

From my expertise as a machine studying researcher, and seeing my algorithms “in action” in several domains, from music platforms to symphonic music live shows, I spotted the large impression that these algorithms have on folks (e.g. listeners, musicians) and directed my analysis towards AI analysis slightly than growth, particularly on learning the impression of AI on human conduct and how one can consider techniques by way of elements comparable to equity, human oversight or transparency. That is my crew’s present analysis matter on the Joint Analysis Centre.

What work are you most pleased with (within the AI subject)?

On the tutorial and technical aspect, I’m pleased with my contributions to music-specific machine studying architectures on the Music Expertise Group in Barcelona, which have superior the state-of-the-art within the subject, because it’s mirrored in my quotation information. As an example, throughout my PhD I proposed a data-driven algorithm to extract tonality from audio indicators (e.g. if a musical piece is in C main or D minor) which has turn into a key reference within the subject, and later I co-designed machine studying strategies for the automated description of music indicators by way of melody (e.g. used to seek for songs by buzzing), tempo or for the modeling of feelings in music. Most of those algorithms are presently built-in into Essentia, an open supply library for audio and music evaluation, description and synthesis and have been exploited in lots of recommender techniques.

I’m significantly pleased with Banda Sonora Very important (LifeSoundTrack), a challenge awarded by Crimson Cross Award for Humanitarian Applied sciences, the place we developed a customized music recommender tailored to senior Alzheimer sufferers. There’s additionally PHENICX, a big European Union (EU)-funded challenge I coordinated on the usage of music; and AI to create enriched symphonic music experiences.

I really like the music computing neighborhood and I used to be completely satisfied to turn into the primary feminine president of the Worldwide Society for Music Info Retrieval, to which I’ve been contributing all my profession, with a particular curiosity in growing variety within the subject.

At present, in my function on the Fee, which I joined in 2018 as lead scientist, I present scientific and technical assist to AI insurance policies developed within the EU, notably the AI Act. From this current work, which is much less seen by way of publications, I’m pleased with my humble technical contributions to the AI Act — I say “humble” as you could guess there are various folks concerned right here! For example, there’s numerous work I contributed to on the harmonization or translation between authorized and technical phrases (e.g. proposing definitions grounded in present literature) and on assessing the sensible implementation of authorized necessities, comparable to transparency or technical documentation for high-risk AI techniques, general-purpose AI fashions and generative AI.

I’m additionally fairly pleased with my crew’s work in supporting the EU AI legal responsibility directive, the place we studied, amongst others, specific traits that make AI techniques inherently dangerous, comparable to lack of causality, opacity, unpredictability or their self- and continuous-learning capabilities, and assessed related difficulties introduced with regards to proving causation.

How do you navigate the challenges of the male-dominated tech business, and, by extension, the male-dominated AI business?

It’s not solely tech — I’m additionally navigating a male-dominated AI analysis and coverage subject! I don’t have a way or a method, because it’s the one surroundings I do know. I don’t know the way it might be to work in a various or a female-dominated working surroundings. “Wouldn’t it be nice?,” just like the Seaside Boys’ music goes. I actually attempt to keep away from frustration and have enjoyable on this difficult state of affairs, working in a world dominated by very assertive guys and having fun with collaborating with glorious girls within the subject.

What recommendation would you give to girls in search of to enter the AI subject?

I’d inform them two issues:

You’re a lot wanted — please enter our subject, as there’s an pressing want for variety of visions, approaches and concepts. As an example, in line with the divinAI challenge — a challenge I co-founded on monitoring variety within the AI subject — solely 23% of writer names on the Worldwide Convention on Machine Studying and 29% on the Worldwide Joint Convention on AI in 2023 have been feminine, no matter their gender identification.

You aren’t alone — there are various girls, nonbinary colleagues and male allies within the subject, regardless that we is probably not so seen or acknowledged. Search for them and get their mentoring and assist! On this context, there are various affinity teams current within the analysis subject. As an example, after I turned president of the Worldwide Society for Music Info Retrieval, I used to be very energetic within the Ladies in Music Info Retrieval initiative, a pioneer in variety efforts in music computing with a really profitable mentoring program.

What are a number of the most urgent points going through AI because it evolves?

In my view, researchers ought to commit as many efforts to AI growth as to AI analysis, as there’s now an absence of stability. The analysis neighborhood is so busy advancing the state-of-the-art by way of AI capabilities and efficiency and so excited to see their algorithms utilized in the actual world that they overlook to do correct evaluations, impression evaluation and exterior audits. The extra clever AI techniques are, the extra clever their evaluations ought to be. The AI analysis subject is under-studied, and that is the reason for many incidents that give AI a nasty fame, e.g. gender or racial biases current in knowledge units or algorithms.

What are some points AI customers ought to concentrate on?

Residents utilizing AI-powered instruments, like chatbots, ought to know that AI just isn’t magic. Synthetic intelligence is a product of human intelligence. They need to be taught in regards to the working ideas and limitations of AI algorithms to have the ability to problem them and use them in a accountable method. It’s additionally vital for residents to be told in regards to the high quality of AI merchandise, how they’re assessed or licensed, in order that they know which of them they will belief.

What’s the easiest way to responsibly construct AI?

In my opinion, the easiest way to develop AI merchandise (with a great social and environmental impression and in a accountable method) is to spend the wanted assets on analysis, evaluation of social impression and mitigation of dangers — for example, to elementary rights — earlier than putting an AI system out there. That is for the advantage of companies and belief on merchandise, but additionally of society.

Accountable AI or reliable AI is a method to construct algorithms the place elements comparable to transparency, equity, human oversight or social and environmental well-being must be addressed from the very starting of the AI design course of. On this sense, the AI Act not solely units the bar for regulating synthetic intelligence worldwide, however it additionally displays the European emphasis on trustworthiness and transparency — enabling innovation whereas defending residents’ rights. This I really feel will improve citizen belief within the product and expertise.