Data is the basis of AI – but how should it be collected?


This is an ad. It is not written by Ny Teknik’s editorial staff.

Artificial intelligence is here for real now. The algorithms are powerful and the possibilities are many. However, data needs to be collected for the training of the algorithms – and it must be done in an ethical way. Something that is easier said than done.

Artificial intelligence, AI, is something that has been talked about for a long time now and over the years, technological development has progressed strongly. The algorithms have become increasingly powerful. The potential for the use of AI is enormous and how the algorithms are used is changing radically.

– It is enough to check social media. Instagram, which has been very good at designing content, has not adapted the algorithms for what we actually like. Instead, we get suggestions for new content based on which we follow. And it can be an old high school friend that we do not share interests with otherwise. At TikTok, the algorithms instead try to learn what you actually want to see, regardless of which ones you follow, says Fredrik Silfver, CTO at Above Agency.

AI has learned to respond

The American company OpenAI has taken it all both one and three steps further. They have created GPT-3, a language model that with in-depth learning has been trained to produce human-like text. In the first generation, the program learned text from across Wikipedia. In generation three, the data came from all over Wikipedia, all available free books and all downloadable material on the internet. From all the data, the program learns the most logical word order and can thereby produce text.

– A journalist who has had to test the program and played around in it has been able to have “conversations” with deceased, famous people and has received well-articulated answers. It is both scary and entertaining, says Fredrik Silfver.

The program can also create advanced artificial images based on photos from the internet. With the new algorithms, a thousand images are no longer required for the AI ​​to learn what, for example, a cat is, but a few images are enough.

The algorithms learn human prejudices

However, the learning itself is not the only challenge with the new technology. For many, there is a goal that AI should be able to reduce discrimination by ignoring factors such as gender and skin color. But, because the data that trains the algorithms comes from humans, the AI ​​also learns our prejudices.

– If we are to be able to achieve an AI that is neutral, and that does not have our faults and shortcomings, we must also wash the data that is fed to the algorithms. I think we still need to figure it out a little more. We will certainly also see a number of new professional roles emerge around AI development, already today there are many who work with training the algorithms not to behave in a discriminatory manner, says Fredrik Silfver.

Read more: How design can be used to tackle ethical challenges in AI.

Ethical data collection required

There is another challenge with data – how should it be collected?

– Our customers ask us for help with how they strategically design software so that it becomes user-friendly and at the same time collects data without infringing on privacy. We then help with advice on the ethical, and on what data they actually need, says Fredrik and concludes:

– Many discussions will be required about how data collection takes place ethically, but the work of collecting data must begin now. This applies not least to large companies. Otherwise, you will be sitting on the lap of Google that has already started.

Facts

Above Agency is an innovation agency with expertise in strategy, design and technology. With a team of over 80 designers and engineers, they create beautiful, smart and valuable products and services for a better tomorrow. Above has employees in Sweden, the USA and China, and customers all over the world. Read more about Above Agency here.

Advertising information

This is an advertisement. It is not written by Ny Teknik’s editorial staff. In case you have comments on the content of the ad, please contact the advertiser.

Do you want to advertise or get in touch with Ny Teknik’s commercial editorial staff? Send an email to agency@nyteknikgroup.se


Releated

Boat and mysterious group named in Nord Stream investigation

SOCIETY Web map of the gas line’s route and Rostock marked where a boat involved in the explosions may have left.Björn Hellström A rented boat that left Rostock, Germany, is suspected of being connected to the explosions of the gas pipelines Nord Stream 1 and 2, according to information in several German media. Prosecutors in […]