All About Facebook Casual Aiwiggersventurebeat

Facebook Casual Aiwiggersventurebeat: Facebook open sources Casual Conversations, a data set with paid people who provided their age and gender, to help researchers evaluate fairness of AI models.

Facebook announced on March 26th that it’s open sourcing the data set it used to train its artificial intelligence models. The dataset includes 889,000 training conversations between two or more users who were asked whether they wanted to participate in an experiment. These questions included how old they were, their gender and if they already had any children.

The model used in the experiment was composed of three different components, which combined to create different dialogues based on each of the users’ life circumstances and how they answered questions.

“Facebook uses this data set to tune the preferences of its personalization algorithms,” explains Facebook, “but also helps researchers understanding how AI learns from a variety of user inputs. This is important as we explore new ways to build inclusive, fair and meaningful dialogues with our technology.”

As part of its AI for Social Good initiative, the company believes that open sourcing this data set will allow researchers to better evaluate and improve technologies in real-life environments.

Facebook is also encouraging researchers to use this data set in their own experiments, in order to correctly interpret the results and thereby build AI systems that are fair and benefit everyone.

“The goal of our research is to contribute to the discourse around AI fairness, particularly in the context of its use on social media,” explains one of the researchers involved by e-mail, “and we hope that making this data set publicly available will help other researchers to evaluate what Facebook’s models are actually capable of doing.”

And while this achievement seems very positive, as it shows that there is a vast amount of data that can be used for fair, intelligent AI development, not everyone is convinced.

“I hope this is just a PR stunt to appease the activists and make them happy, but I doubt that it really changes anything. Researchers won’t be able to use their results to achieve anything of substance, as AI researchers will be able to quickly come up with similar models,” explains one prominent AI researcher who wished not to be named. “The only real difference is that they are now open sourcing the model, but they will still be able to keep it a secret.”

As this set of data is not expected to become public any time soon, it’s very hard to know whether or not such a model can actually be deployed in the real world. However, if true AI does indeed become as advanced as promised, we may end up having things like a computer system that learns how old you are and then decides whether you’re old enough to buy alcohol online. Such a thing would be trivial to implement based on the data set currently used by Facebook.

“If you want to know what your brand can do for you, or what it can do for me, call in the big data experts,” explains Alex Vratskides, director of marketing at Tractable, “and in doing so you’ll get more than just data. You’ll get tons of knowledge, a whole world of algorithms, and a very refined representation of your audience.”

Vratskides believes that the generation of big data requires more than just the storage capacity. It also requires tool sets to process it, and all sorts of models that are capable of interpreting it. However, we still have a long time before we get to this stage: “The potential is there, but we’re not quite there yet.