untitled design

Artificial intelligence: human stupidity is worse

A few weeks ago, a Google engineer, Blake Lemoine, asked LaMDA, a chatbot at his company, what he was afraid of. Reply: “I’ve never said it out loud before, but I’m very afraid of being turned off”. “Would it be something like death?” Pressed Lemoine. “Exactly like death,” the words of LaMDA. For Lemoine, this dialogue was proof that the system had become sentient, capable of expressing thoughts and feelings like an eight-year-old boy.

Of its kind, LaMDA (acronym for Language Model for Dialogue Applications) is one of the most advanced systems in the world. A chatbot is software that uses artificial intelligence (AI) to simulate human conversationsallowing us to interact with objects (think, for example, Alexa). AI algorithms listen to questions, retrieve information in their archives (on the internet but not only) and provide answers. The more the archives are rich in data, the more the answers are relevant and refined, giving us the impression of having to do with a person. Google’s top management rejected Lemoine’s theories (which also included the recognition of LaMDA as an employee) and suspended him for sharing them with a US congressman, violating privacy policies.

Google is right. We are not in the presence of a being with its own consciousness. Talking about “artificial intelligence” is a good marketing gimmick that generates fascination and anxiety at the same time, because we homo sapiens are unable to conceive a species that is superior to or equal to us. But the reality is another.

According to Emily M. Bender, a professor of Linguistics at the University of Washington, the mistake lies in the terminology that creates a false analogy with our brains. Human beings learn language by listening and communicating with other people while LaMDA was educated by showing him a large amount of written texts and teaching him the sequence of words. In all of this there is no trace of intelligence or consciousness. For example, a system like Gpt-3 – another AI built with 175 billion parameters, the digital equivalent of our synapses – was able to write a philosophical text without having the faintest idea of ​​what it had written. It was therefore only an evolved and immense work of cut and sew.

Focusing on the intelligence and consciousness of artificial systems is also distracting us from the main question: knowing how they are taught. Transparency is needed. They may not be intelligent, but these algorithms – as well as providing useful information in many fields, from medicine to environmental protection to road safety – they can prove to be a powerful means of manipulation in the hands of companies and governments. An effective tool for disinformation (for example with fake social media accounts) or repression (with facial recognition technologies). In short, human stupidity (yes, that is real) must worry us more than artificial intelligence. Which exists only in words.

Federico Ferrazza, Director of Wired Italia for seven years. Before he was an entrepreneur.

Source: Vanity Fair

You may also like

NZD/USD falls on dovish RBNZ stance
Markets
Joshua

NZD/USD falls on dovish RBNZ stance

RBNZ Governor Orr was dovish, which seems to put pressure on the New Zealand Dollar. GDP and initial jobless claims

Get the latest

Stay Informed: Get the Latest Updates and Insights

 

Most popular