“Artificial intelligence is moving the boundary between human and non-human”
[ad_1]
Alexei Grinbaum is a physicist and philosopher at the French Atomic Energy and Renewable Energies Commission (CEA). A member of the National Pilot Committee for Digital Ethics, he alerts us to the progress, both fascinating and worrying, in the field of automatic language generation. In particular, it returns for The world on employee statements of Google, Blake Lemoine, in the washington post from June 11. After talking with the company’s latest natural language processing algorithm, LaMDA, this engineer explained that he believed he was talking to a human intelligence to which he attributes the sensitivity of a 7-8 year old child. The revelation of the details of the conversations revealed the considerable progress of Google in terms of language, and revived the debates on the capacities of the machines. For having, according to the company, violated its safety policy, Blake Lemoine was dismissed on July 22.
How does this realistic conversation with artificial intelligence inspire you?
We know since 1965, and the first chatbot [agent conversationnel], Eliza, that talking machines can have psychological effects on users. Eliza was a deterministic, fairly trivial system whose strategy was to turn assertions into questions to get the conversation going again. With current artificial intelligence (AI) systems, the strategy is different, but Blake Lemoine also suffers effects: he is under the illusion that this machine has emotions, moods… However, knowing that he is a machine does not prevent him from making, very seriously, projections of human qualities on LaMDA. It projects feelings, as we sometimes do with our beloved car or our mischievous smartphone…
“LaMDA approaches the human because it also creates the illusion of having free will”
The illusion is still almost perfect, isn’t it?
The particularity is that this illusion comes from language, without the need to resort to a pretty humanoid figure, such as the sophia-robot. TheMDA writes text, and this is enough for the projection of human features. It should be remembered that in the Middle Ages, a “monster” could be baptized if it fulfilled one of three criteria: human paternity, human face or language. As soon as a strange entity speaks in our language, a relationship is automatically established with the human interlocutor. In the myths, the non-human agents who spoke to us were gods, demons or angels. From now on, we interact with technical prowess, talking machines.
You have 75.59% of this article left to read. The following is for subscribers only.
[ad_2]
Source link
