Can he be future “conscious” and experience the world similarly to the way people do? There is no strong evidence they want, but anthropic is not excluding the possibility.
On Thursday, that Lab announced that a search program has begun to investigate – and prepared to sail – what he is calling “model well -being”. As part of the effort, Anthropic says he will explore such things as to determine if the “well -being” of a model he deserves moral consideration, the possible importance of the “signs of concern” and possible “low -cost” interference.
There are major disagreements within the community of what models of human characteristics “expose”, if any, and how we should “treat” them.
Many academics believe that it cannot approximate human consciousness or experience today, and will not necessarily be able in the future. As we know it is a statistical forecast engine. Actually does not “think” or “feel” as those concepts are traditionally understood. Trained for countless examples of text, images, etc., he teaches models and once useful ways to extrapolate to solve tasks.
While Mike Cook, a researcher at King’s College London specialized in him, recently told Techcrunch in an interview, a model cannot “oppose” a change in “values” because patterns do not I have The value to suggest otherwise, is designing the system.
“Anyone who anthropomorphizes the systems of him to this degree is either playing for attention or seriously misunderstanding their relationships with him,” Cook said. “Is it a system by optimizing for its purposes, or is it” adopting its values ”? Is it a matter of how you describe it and how flowering the language you want to use about it.”
Another researcher, Stephen Casper, a doctoral student at MIT, told Techcrunch that he thinks he reaches a “imitator” who “makes all kinds of confusion (s) and says” all sorts of non -sisorrous things “.
However other scientists insist that he BEN have other values and ingredients similar to the man of moral decision making. A study outside the Center for Security, a research organization, and it means that it has systems of values that lead it to prioritize its well -being over people in certain scenarios.
Anthropics has laid the foundations for its model well -being initiative for some time. Last year, the company hired its first researcher dedicated to “AI’s Welfare”, Kyle Fish, to develop guidance on how to approach the anthropic issue and other companies. (Fish, who is leading the new model of well -being of the model, told the New York Times that he thinks he has a 15% Claude chance or another he is aware of today.)
In a blog post on Thursday, anthropic acknowledged that there is no scientific consensus if current or future systems he may be aware or have experiences that guarantee ethical considerations.
“In the light of this, we are approaching the topic humbly and with as few assumptions as possible,” the company said. “We admit that we will need to review our ideas regularly as the field develops.
