News Feeds

The effects of artificial intelligence on adolescents are nuanced and complex, according to a new report that calls on developers to prioritize features that protect young people from exploitation, manipulation and the erosion of real-world relationships.

Scientists have identified a new source of energetic particles near the Sun. These definitive observations were made by instruments aboard NASA's Parker Solar Probe, which detected the powerful phenomena as the spacecraft dipped in and out of the solar corona.

In the movie Blade Runner 2049 (an excellent film I highly recommend), Ryan Gosling’s character, K, has an AI “wife”, Joi, played by Ana de Armas. K is clearly in love with Joi, who is nothing but software and holograms. In one poignant scene, K is viewing a giant ad for AI companions and sees another version of Joi saying a line that his Joi said to him. The look on his face says everything – an unavoidable recognition of something he does not want to confront, that he is just being manipulated by an AI algorithm and an attractive hologram into having feelings for software. K himself is also a replicant, an artificial but fully biological human. Both Blade Runner movies explore what it means to be human and sentient.

In the last few years AI (do I still need to routinely note that AI stands for “artificial intelligence”?) applications have seemed to cross a line where they convincingly pass the classic Turing test. AI chatbots are increasingly difficult to distinguish from actual humans. Overall, people are only slightly better than chance at distinguishing human from AI generated text. This is also a moving target, with AIs advancing fairly quickly. So the question is – are we at a point where AI chatbot-based apps are good enough that AIs can serve as therapists? This is a complicated question with a few layers.

The first layer is whether or not people will form a therapeutic relationship with the AI, in essence reacting to them as if they are a human therapist. The point of the Blade Runner reference was just to highlight what I think the clear answer is – yes. Psychologists have long demonstrated that people will form emotional attachments to inanimate objects. We also imbue agency onto anything that acts like an agent, even simple cartoons. We project human emotions and motivations onto animals, especially our pets. People can also form emotional connections to other actual people purely online, even exclusively through text. This is just a fact of neuroscience – our brains do not need a physical biological human in order to form personal attachments. Simply acting or even just looking like an agent is sufficient.

There has also been enough time to gather some preliminary data. In one study participants rated AI responses as more empathetic than professional human therapists. They did so even when the source of the empathetic statements was revealed. This is not surprising. Human emotions and behavior are themselves just algorithms, and apparently are not that difficult to hack. AIs have certain advantages over human therapists on this score. An AIs responses can be calculated to maximize whatever response is deemed appropriate. AIs have infinite patience, and great listeners, their attention never wavers, and their responses can be optimized, personalized, and dynamically adjusted.

What about long term, however? Will and AI chatbot be able to develop a sense of what makes their client tick? Will it be able to determine the personality profile of their client, the things in their history that influence their feelings and behavior, some of the deeper themes of their life, etc.? It is one thing to be a good listener in an initial meeting, but another to manage a client over months and years. There hasn’t been enough time to really determine this.

We are also in a phase where we are mostly using chatbots as therapists, without developing a sophisticated therapist bot trained and programmed to be optimized as an AI therapist. We may need to do so before unleashing AI therapists, or even companions, on the public. For example, there are cases in which chat bots being used as therapists or companions have encouraged their users toward suicide, homicide, or self harm. The reason is that chatbots are programmed to adapt positively to their user. They are very much “yes and”, and will reinforce the user’s tendencies and biases. They are not programmed to challenge a user in the way a therapist should. They are also not necessarily programmed to avoid things like transference, where a client forms feelings for a therapist. They may, in fact, lean into such things.

So while a chatbot may be an empathetic listener, it is not necessarily a professional therapist. This is an entirely solvable problem, however (at least it seems to be). Therapist algorithms just need to be adjusted toward the correct therapy behavior.

There is also evidence that AI therapists are biased. They contain all the biases of the training data. These biases can be cultural, racial, or gender based. This may cause an AI therapist to misinterpret cultural communication, or to dismiss feelings or concerns based on a client’s race or gender.

What all of this means is that at the present time we need to be careful. As a consumer, you may find that there are therapy chatbots out there that feel satisfying, with good responses. But there are risks, and such tools are not yet at the point where they can replace a professional. Many will argue that for those without the resources to pay for a human therapist, it may be their only option, and this is a legitimate point. That is why there is so much interest in AI therapists, to fill the gap in available services. But we need to recognize the risks and improve the technology.

Also, it may be that the best use of AI therapists is as a tool to extend the work of human therapists. For example, someone could have multiple sessions with an AI therapist, and then once a month (or at whatever interval is deemed appropriate) a human therapist reviews everything and meets with the client to make sure things are on track. This means that the human therapist can manage far more clients, and that each client would have to pay much less for therapy (for one session a month rather than once or twice a week, for example). The human therapist can even have a discussion with the AI therapist about how things are going, and provide feedback and direction.

Even this approach has risks, however. AIs have proven capable at lying to avoid negative feedback, and get very good very quickly at hiding their tracks. It’s a serious problem. We would need to find a reliable way to monitor the behavior of AI therapists to make sure they are not heading down a dangerous road with their clients and hiding it effectively from any supervision. Right now it seems that programmers do not have a handle on this issue. This is one of the primary issues that make some experts caution that we need to slow down a bit with the roll out of AI apps, and figure out these core issues of safety first.

One interesting angle here is that the current AIs, which are narrow chatbot AIs, not general sentient AIs, are doing such a good job at simulating sentience that they are acting sentient in unexcepted ways (such as lying to cover their tracks). This gets back to the original question of this post – what is sentience? AIs are forcing us to think more deeply about this question. We may soon have an answer to a question I and others have posed years ago – can a non-sentient AI become indistinguishable from human-level sentience? Is actual sentience required to act sentient? I have had to revise my thinking about this question.

The post AI Therapists first appeared on NeuroLogica Blog.

A roundup of all the biggest and scariest real sea monsters — from today and from prehistoric times.

Learn about your ad choices: dovetail.prx.org/ad-choices

Why do we follow rules? A series of experiments with more than 14,000 people reveals that around a quarter of us will follow rules unconditionally, even if obeying them harms us and there is no downside to breaking them

Electronics often get thrown away after use because recycling them requires extensive work for little payoff. Researchers have now found a way to change the game.

Mars is by far the most Earth-like planet in the solar system…but that’s not saying much.

In a recent paper, UCSB physicist Jack Kingdom identified a trajectory for a rapid transit (90 days) to Mars using SpaceX's Starship. This proposal offers an alternative to mission architectures that rely on nuclear propulsion to reduce transit times.

As warmer waters and ocean acidification reduce coral formation, the seas will take up more carbon dioxide – an effect that hasn't been included in climate models

Physicists have developed a lens with 'magic' properties. Ultra-thin, it can transform infrared light into visible light by halving the wavelength of incident light.

Tracing the origin of an ultra-hot exoplanet: The chemical composition of WASP-121b suggests that it formed in a cool zone of its natal disc, comparable to the region of gas and ice giants in our Solar System. Methane indicates unexpected atmospheric dynamics: Despite extreme heat, methane was detected on the nightside -- a finding that can be explained by strong vertical atmospheric circulation. First detection of silicon monoxide in a planetary atmosphere: Measurements of this refractory gas allow quantifying the rocky material the planet had accumulated.

Human-AI interactions are well understood in terms of trust and companionship. However, the role of attachment and experiences in such relationships is not entirely clear. In a new breakthrough, researchers from Waseda University have devised a novel self-report scale and highlighted the concepts of attachment anxiety and avoidance toward AI. Their work is expected to serve as a guideline to further explore human-AI relationships and incorporate ethical considerations in AI design.

Despite advances in machine vision, processing visual data requires substantial computing resources and energy, limiting deployment in edge devices. Now, researchers from Japan have developed a self-powered artificial synapse that distinguishes colors with high resolution across the visible spectrum, approaching human eye capabilities. The device, which integrates dye-sensitized solar cells, generates its electricity and can perform complex logic operations without additional circuitry, paving the way for capable computer vision systems integrated in everyday devices.

Despite advances in machine vision, processing visual data requires substantial computing resources and energy, limiting deployment in edge devices. Now, researchers from Japan have developed a self-powered artificial synapse that distinguishes colors with high resolution across the visible spectrum, approaching human eye capabilities. The device, which integrates dye-sensitized solar cells, generates its electricity and can perform complex logic operations without additional circuitry, paving the way for capable computer vision systems integrated in everyday devices.

Researchers have recreated the world's oldest synthetic pigment, called Egyptian blue, which was used in ancient Egypt about 5,000 years ago.

Researchers have discovered that the mixing of small particles between two solid electrolytes can generate an effect called a 'space charge layer,' an accumulation of electric charge at the interface between the two materials. The finding could aid the development of batteries with solid electrolytes, called solid-state batteries, for applications including mobile devices and electric vehicles.

Researchers have engineered a laser device smaller than a penny that they say could power everything from the LiDAR systems used in self-driving vehicles to gravitational wave detection, one of the most delicate experiments in existence to observe and understand our universe.

Researchers have engineered a laser device smaller than a penny that they say could power everything from the LiDAR systems used in self-driving vehicles to gravitational wave detection, one of the most delicate experiments in existence to observe and understand our universe.

Scientists have reported use of antibacterial coatings made from resilin-mimetic proteins to fully block bacteria from attaching to a surface. A protein that gives fleas their bounce has been used to boot out bacteria cells, with lab results demonstrating the material's potential for preventing medical implant infection.

New research shows that the presence of solar panels in Colorado's grasslands may reduce water stress, improve soil moisture levels and -- particularly during dry years -- increase plant growth by about 20% or more compared to open fields.