Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Get the latest international news and world events from around the world.

Log in for authorized contributors

No, Artificial Intelligence Is Not Conscious

Earlier this year, Anthropic released a “constitution” for Claude, its large language model and flagship product; Anthropic CEO Dario Amodei has said “we’re open to the idea” that AI could be conscious; and Anthropic’s in-house philosopher, Amanda Askell, said in an interview, “I want Claude to be very happy.”

“It’s enough to make you wonder: Should we seriously consider the possibility that Claude, or any large language model, might be conscious? And if it has feelings, is it capable of receiving moral instruction?” Ted Chiang asks. “Absolutely not.”

“LLM conversations are cleverly disguised examples of sentence continuation,” Chiang writes. Perhaps the most fruitful way to understand Claude’s constitution “is as an 83-page character sheet for a role-playing game. LLMs can generate dialogue for Julius Caesar because many books about him exist in the training data those models used. Claude’s constitution serves a similar role for delineating the helpful-chatbot character that customers interact with when they’re using Anthropic’s products.”

“The result is a sentence-continuation machine that is likelier to emit sentences resembling those that a thoughtful, moral person could utter,” Chiang continues. “However, for all the times that ‘honesty’ is mentioned in Claude’s constitution, I would argue that it is fundamentally dishonest to have a machine emit many categories of sentences including any sentences using first-person pronouns.”

“Whenever a person delegates a decision to an LLM, they are trying to off-load accountability for that decision, and if a company that sells an LLM portrays the product as having a moral center, it is offering a way for its customers to abdicate their responsibilities,” Chiang writes. “Off-loading tasks such as writing code might result in cognitive atrophy over the long term, and that is problematic in itself, but off-loading ethical decisions will result in an atrophy of moral reasoning, which is worse.”

“It’s fortunate that LLMs are not conscious,” Chiang continues, “or else the actions of the big AI firms would be even more scandalous than they already are.”

Continue reading “No, Artificial Intelligence Is Not Conscious” | >

New brain wave theory explains cognition and consciousness

A new theory, published in The Journal of Neuroscience by three scientists in The Picower Institute for Learning and Memory at MIT, offers an explanation of how the brain produces cognition and consciousness: It uses traveling waves of rhythmic neural activity to coordinate nimble neural networks with analog computations.

The metaphor that the brain operates with “circuits” is incomplete, said Picower Professor Earl K. Miller, the paper’s senior author. Indubitably, the brain’s physically connected circuits provide the infrastructure to store our memories and represent our ongoing needs and goals. But when we need to make improvised use of that knowledge in the rapid-fire, anything goes sensory context the world constantly throws our way, we can’t just depend on the relatively slow chemical process of rewiring those circuit connections called “synapses,” he said. Instead, the brain needs a control system that can coordinate millions of neurons to process information in a fraction of a second. Brain waves, long understood to be the synchronized rhythmic fluctuations of large groups of neurons, turn out to be performing that crucial service, Miller and his colleagues argue, citing years of experimental evidence from his lab and many others.

Circuits and synapses are important and fundamental, that’s the start. But there is more going on. The brain generates waves, and wave dynamics are a highly efficient way to coordinate and perform computation

Computer models pinpoint catalysts for replacing fossil-fueled ammonia production

Ammonia is one of the most important chemicals produced in the world, ranking second only to sulfuric acid in the total volume produced each year. It is used mostly to make fertilizer, which is essential to feeding the world’s population. Yet its production accounts for up to 2% of the world’s energy consumption and about 1.5% of greenhouse gas emissions, so the search has been underway for ways to produce ammonia more sustainably.

The traditional way of making ammonia, in use for more than a century and accounting for the vast majority of production, is the Haber–Bosch process, which relies on fossil fuels to provide the needed heat. Hydrogen used in the process is also largely produced from fossil fuels.

There is another way, using electrochemistry instead of heat and pressure, but so far this method has not been anywhere near economically competitive at the scales needed.

First-in-class therapy targets ‘undruggable’ protein in hard-to-treat blood cancers

A first-in-class therapy to target MYC, one of the most sought-after and difficult targets in cancer biology, showed promise in hard-to-treat blood cancers, according to a new preclinical study from The University of Texas MD Anderson Cancer Center published in Blood.

Researchers led by Michael Andreeff, M.D., Ph.D., professor, and Yuki Nishida, M.D., Ph.D., assistant professor, both of Leukemia, found that the experimental drug GT19630 interrupts a newly discovered cycle between MYC and GSPT1, resulting in strong anticancer activity in preclinical models of leukemia, lymphoma and multiple myeloma, including treatment-resistant and TP53-mutated disease.

“For decades, scientists have struggled to develop therapies that successfully block MYC, leading many in the field to describe it as undruggable,” Andreeff said. “By identifying a vulnerability in the relationship between MYC and GSPT1, we found a way to eliminate both proteins and disable a pathway many cancers depend on for survival.”

The silent data leak hidden inside encrypted reasoning traces of frontier AI models

To hide this reasoning while maintaining conversational context across multiple API turns, major providers such as Anthropic and OpenAI use Authenticated Encryption with Associated Data (AEAD) envelopes. AEAD is a cryptographic standard that encrypts data while attaching metadata to verify that the payload has not been modified in transit.

Providers designed these encrypted envelopes around three goals:

1. Confidentiality: Prevent clients from reading the intermediate reasoning tokens to protect proprietary information.

The dangers of trusting blackbox machine learning

The success of deep learning in the past decade has increased interest in the field of artificial intelligence. But the rising popularity of AI has also highlighted some of the key problems of the field, including the “black box problem,” the challenge of making sense of the way complex machine learning algorithms make decisions. The Apple Card disaster is one of many manifestations of the black-box problem coming to light in the past years.

The increased attention to black-box machine learning has given rise to a body of research on explainable AI. And a lot of the work done in the field involves developing techniques that try to explain the decision made by a machine learning algorithm without breaking open the black box. But explaining AI decisions after they happen can have dangerous implications, argues Cynthia Rudin, professor of computer science at Duke University, in a paper published in the Nature Machine Intelligence journal.

“Rather than trying to create models that are inherently interpretable, there has been a recent explosion of work on ‘explainable ML’, where a second (post hoc) model is created to explain the first black box model. This is problematic. Explanations are often not reliable,” Rudin writes. and can be misleading, as we discuss below.

/* */