We inform you that on this website we use our own and third-party cookies to collect information about its use, improve our services and, where appropriate, display advertising by analyzing your browsing habits. You can expressly accept its use by pressing the "ACCEPT" button or configure and select the cookies you want to accept or reject in the settings. You can also get more information about our cookie policy here.

The global fashion business journal

Dec 5, 20247:51pm

Michael Veale (University College of London): “Design still needs human intervention”

The specialized researcher in fairness, transparency and resilience of algorithmic systems addressed the challenges of machine learning in fashion during the first edition of Shaking Fashion.

Feb 15, 2019 — 9:59am
Iria P. Gestal
Save

Michael Veale (University College of London): “Design still needs human intervention”

 

 

Where machines fail, a human will still be needed. Those were the words of Michael Veale, researcher at the University College of London and in charge of giving the kick-off to the Shaking Fashion series of meetings, organized by Modaes.es and EY. The expert defends that, in fashion, the human being will continue to be necessary in design and that, although machine learning has great possibilities for the sector, the challenge lies in the management of the data and the predictive capacity of the machines.

 

“Machine learning has two barriers: on the one hand, it can only predict from what has already happened, so if traditionally there were inequalities or discrimination, it will replicate it,” noted Veale.

 

“On the other hand, the human being understands causality, while the machine understands only correlation, that is why it is still necessary to sophisticate further technology,” defended the expert, who stressed that “design will still need human intervention.”

 

 

 

 

Veale researches on equity, transparency and resilience of algorithmic systems and is completing a PhD in the Department of Science, Technology, Engineering and Public Policy at the University College of London. The researcher has worked on the development of algorithmic decision-making systems and analysis of online content moderation systems.

 

In the fashion industry, the main uses of artificial intelligence and machine learning are aimed at the prediction of habits and tastes through data analysis, but this presents great challenges, such as the management of data protection and errors in the prediction.


“The consumer is resigned to his safety: it is a complicated topic,” affirmed Veale. Author of the report Data Management and Use: Governance in the 21st Century, and ex-adviser of the European Commission in health governance and the Internet of Things.

 

 

 

 

At all events, the biggest problem lies not so much in obtaining such data but in making sure that they resemble reality. That is why Veale distrusts the utility of technologies such as blockchain: “it is not a useful tool”.

 

The academic gave as an example the human resources algorithm that Amazon had to withdraw because it discriminated against the offers of female costumers. “We must think about this kind of risk since the design process,” defended Veale. “A scanner to predict sizes when buying online, for example, is a good tool, but what happens if it does not detect someone who wears a size above 38? Or what about someone with some type of disability? Your data will not be correct, and therefore the predictions will not either”.

 

Veale also reviewed the different specific applications of machine learning that are already being used in the field of fashion. Most are applied to the online environment, where the ability to collect and analyse data is much greater.

 

In addition to scanning and creating body models to assist online shopping, artificial intelligence also allows, for example, automatic cataloguing of garments from the information sent by the supplier and the image of the item, a system that is already applied on platforms like Amazon.

 

 

 

 

Another application of machine learning in fashion is the search for a product from an image, or even by modifying the content of a photograph: taking, for example, an image of a green dress, and looking in a website that same dress, but sleeveless or shorter.

Finally, this technology also allows to perform market studies using the large amount of images that people published every day on the Internet to identify what are the most popular items or styles.

 

In the medium term, other uses may be added, such as the so-called affective computing or emotional AI, which will predict the mood of a person and even reflect it in their clothes, with a very similar technology to the one smartwatches already use.

 

Shaking Fashion is a cycle of meetings promoted by Modaes.es and EY with which we want to invite entrepreneurs and managers of the fashion business in Spain to make a shake with experts who usually do not share their knowledge to look beyond the present.    


Advertising
Participation rules

info@themds.com

 

Validation policy for comments: 

 
MDS does not perform prior verification for the publication of comments. However, to prevent anonymous comments from affecting the rights of third parties without the ability to reply, all comments require a valid email address, which won’t be visible or shared.
 
Enter your name and email address to be able to comment on this news: once you click on the link you will find within your verification email, your comment will be published.

0 comments — Be the first to comment
...