Senator Chris Murphy’s ChatGPT tweet is proof lawmakers aren’t ready for the AI ​​boom

On Sunday night, Sen. Chris Murphy (D-CT) tweeted a bold claim about ChatGPT, saying the chatbot “taught itself to do advanced chemistry” even though chemistry knowledge was not “built into the model” and no one “programmed it on it.” learn complicated chemistry.”

“It decided to teach itself and then made its knowledge available to anyone who asked,” Murphy added. “Something’s coming. We are not ready.”

The only problem: Almost everything Murphy wrote in that tweet was wrong. AI researchers and academics were quick to let him know, flooding his replies and quoting tweets with the kind of righteous tone and fury reserved for the internet’s protagonist of the era.

“It’s just not true,” says Grady Booch, software engineer and developer of the Unified Modeling Language. wrote in Murphy’s replies. “Please, you need to educate yourself better about the reality of contemporary AI.”

“Your description of ChatGPT is dangerously misinformed,” Melanie Mitchell, AI researcher and professor at the Santa Fe Institute, wrote back in another tweet. “Every sentence is wrong. I hope you will learn more about how this system actually works, how it was trained and what its limitations are.”

Aside from being a great example of something that should have stayed in the draft folder, Murphy’s tweet underscores the stark reality that the vast majority of our lawmakers are woefully unprepared for the AI ​​boom. Since the release of ChatGPT in November 2022, we’ve seen big tech giants like Microsoft, Google, and China’s Baidu rush to bring generative AI products to market – with varying degrees of success. Microsoft released a new version of Bing enriched with GPT-4, which so scandalized one journalist that he wrote a front-page article about it for the New York Times. The company later unveiled a slew of AI-powered updates to its existing products, like Excel and Word. Meanwhile, Google has caught up, releasing its AI chatbot Bard a month later.

Amidst all this excitement, misinformation about Generative AI is rapidly growing out of proportion. It has led people to fundamental misunderstandings of technology and its capabilities. We see people making outlandish claims like the Bing chatbot has fallen in love with them (it’s not), or that it’s sentient (it’s not), or that it’s evil and wants to kill them (it’s not and it habit).

Now we have an incumbent US Senator with a massive platform adding fuel to that fire. To his credit he did it reply later and seemed to imply that he might have been wrong (or plain wrong) in his first take. A source with intimate knowledge of the situation told The Daily Beast that the source of Murphy’s tweet was a Center for Humane Technology presentation on AI by Aza Raskin and Tristan Harris. Still, it doesn’t make its initial attitude any less wrong — or dangerous.

First, ChatGPT is built using OpenAI’s Large Language Models (LLM) GPT-3.4 and GPT-4. That means it uses a dataset drawn from a huge corpus of books, scholarly journals and articles from various internet sources such as Wikipedia or news websites. This represents literally petabytes of data for the purpose of text prediction. So it can’t and won’t “teach itself” advanced chemistry – or really anything – because it’s a predictive text bot like the one on your phone. It generates responses based on prompts and the words that are likely to follow each other.

“ChatGPT doesn’t teach itself,” Mitchell told The Daily Beast in an email. “People give him huge amounts of text. It is trained to predict the next token in a block of text.”

Mitchell added that while the training allows him to learn what human language looks like, it doesn’t give him the ability to “understand the questions people give him, or the language that speaks it in a human-like way.” generated”.

Moreover, all of this is actually built into the model. That is the point. ChatGPT has been trained to be an incredibly sophisticated and advanced chatbot. “ChatGPT doesn’t decide at all,” Mitchell explained. It has no intentions.”

Mitchell and other AI experts’ frustration is fueled in part by the danger posed by misinformation surrounding these chatbots. When people start treating these bots as these omnipotent or omniscient things, then they’re going to start giving them a level of authority they just shouldn’t have.

“I want Sen. Murphy and other policymakers to know that they pose a huge risk to our information ecosystem,” Emily M. Bender, a professor of linguistics at the University of Washington, told The Daily Beast in an email. “These are programs for creating texts that sound plausible, but have no obligation to the truth.”

She added, “This means our information ecosystem could quickly become swamped with non-information, making it harder to find and trust trusted sources of information.”

Booch largely echoed the sentiment. “Facts matter, and the senator is doing his community and the domain of AI a disservice by spreading misinformation like this,” Booch told The Daily Beast. However, he pointed out that “OpenAI is behaving in a highly unethical manner by not disclosing the source of their corpus”.

There is currently little standing in the way of sensible regulation in the area of ​​AI. In October 2022, the White House introduced a framework for the AI ​​Bill of Rights that outlines principles for how these models can be built and used to protect the data and privacy of American citizens. However, it is currently little more than a glorified wish-list of vague regulation. Since its release, the world of Generative AI has exploded – and with it the risks.

However, Murphy got one thing right: something is coming and we’re not ready. He probably didn’t realize that he was also talking about himself.

“We urgently need intelligent regulation around the collection and use of data, around automated decision-making systems and around accountability for synthetic text and images,” said Bender. “But the people who sell these systems (especially OpenAI) would rather alarm policymakers about doomsday scenarios with mood machines.”