We had a fascinating discussion on this week’s SGU that I wanted to bring here – the subject of artificial intelligence programs (AI), specifically large language models (LLMs), lying. The starting point for the discussion was this study, which looked at punishing LLMs as a method of inhibiting their lying. What fascinated me the most is the potential analogy to neuroscience – are these LLMs behaving like people?

LLMs use neural networks (specifically a transformer model) which mimic to some extent the logic of information processing used in mammalian brains. The important bit is that they can be trained, with the network adjusting to the training data in order to achieve some preset goal. LLMs are generally trained on massive sets of data (such as the internet), and are quite good at mimicking human language, and even works of art, sound, and video. But anyone with any experience using this latest crop of AI has experienced AI “hallucinations”. In short – LLMs can make stuff up. This is a significant problem and limits their reliability.

There is also a related problem. Hallucinations result from the LLM finding patterns, and some patterns are illusory. The LLM essentially makes the incorrect inference from limited data. This is the AI version of an optical illusion. They had a reason in the training data for thinking their false claim was true, but it isn’t. (I am using terms like “thinking” here metaphorically, so don’t take it too literally. These LLMs are not sentient.) But sometimes LLMs don’t inadvertently hallucinate, they deliberately lie. It’s hard not to keep using these metaphors, but what I mean is that the LLM was not fooled by inferential information, it created a false claim as a way to achieve its goal. Why would it do this?

Well, one method of training is to reward the LLM when it gets the right answer. This reward can be provided by a human – checking a box when the LLM gives a correct answer. But this can be time consuming, so they have build self-rewarding language models. Essentially you have a separate algorithm which assessed the output and reward the desired outcome. So, in essence, the goal of the LLM is not to produce the correct answer, but to get the reward. So if you tell the LLM to solve a particular problem, it may find (by exploring the potential solution space) that the most efficient way to obtain the reward is to lie – to say it has solved the problem when it has not. How do we keep it from doing this.

Continue Reading »

If you think about the human hand as a work of engineering, it is absolutely incredible. The level of fine motor control is extreme. It is responsive and precise. It has robust sensory feedback. It combines both rigid and soft components, so that it is able to grip and lift heavy objects and also cradle and manipulate soft or delicate objects. Trying to replicate this functionality with modern robotics have been challenging, to say the least. But engineers are making steady incremental progress.

I like to check it on how the technology is developing, especially when there appears to be a significant advance. There are two basic applications for robotic hands – for robots and for prosthetics for people who have lost their hand to disease or injury. For the latter we need not only advances in the robotics of the hand itself, but also in the brain-machine interface that controls the hand. Over the years we have seen improvements in this control, using implanted brain electrodes, scalp surface electrodes, and muscle electrodes.

We have also seen the incorporation of sensory feedback, which greatly enhances control. Without this feedback, users have to look at the limb they are trying to control. With sensory feedback, they don’t have to look at it, overall control is enhanced, and the robotic limb feels much more natural. Another recent addition to this technology has been the incorporation of AI, to enhance the learning of the system during training. The software that translates the electrical signals from the user into desired robotic movements is much faster and more accurate than without AI algorithms.

A team at Johns Hopkins is trying to take the robotic hand to the next level – A natural biomimetic prosthetic hand with neuromorphic tactile sensing for precise and compliant grasping. They are specifically trying to mimic a human hand, which is a good approach. Why second-guess millions of years of evolutionary tinkering? They call their system a “hybrid” robotic hand because it incorporates both rigid and soft components. Robotic hands with rigid parts can be strong, but have difficulty handling soft or delicate objects. Hands made of soft parts are good for soft objects, but tend to be weak. The hybrid approach makes sense, and mimics a human hand with internal bones covered in muscles and then soft skin.  Continue Reading »

Remember CRISPR (clustered regularly interspaced short palindromic repeats) – that new gene-editing system which is faster and cheaper than anything that came before it? CRISPR is derived from bacterial systems which uses guide RNA to target a specific sequence on a DNA strand. It is coupled with a Cas (CRISPR Associated) protein which can do things like cleave the DNA at the targeted location. We are really just at the beginning of exploring the potential of this new system, in both research and therapeutics.

Well – we may already have something better than CSRISP: TIGR-Tas. This is also an RNA-based system for targeting specific sequences of DNA and delivering a TIGR associated protein to perform a specific function. TIGR (Tandem Interspaced Guide RNA) may have some useful advantages of CRISPR.

As presented in a new paper, TIGR is actually a family of gene editing systems. It was discovered not by happy accident, but by specifically looking for it. As the paper details: “through iterative structural and sequence homology-based mining starting with a guide RNA-interaction domain of Cas9”. This means they started with Cas9 and then trolled through the vast database of phage and parasitic bacteria for similar sequences. They found what they were looking for – another family of RNA-guided gene editing systems.

Like CRISPR, TIGR is an RNA guided system, and has a modular structure. Different Tas proteins can be coupled with the TIGR to perform different actions at the targeted site. But there are several potential advantages for TIGR over CRISPR. Like CRISPR it is RNA guided, but TIGR uses both strands of the DNA to find its target sequence. This “dual guided” approach may lead to fewer off-target errors. While CRISPR works very well, there is a trade-off in CRISPR systems between speed and precision. The faster it works, the greater the number of off-target actions – like cleaving the DNA in the wrong place. The hope is that TIGR will make fewer off-target mistakes because of better targeting.

Continue Reading »

Small nuclear reactors have been around since the 1950s. They mostly have been used in military ships, like aircraft carriers and submarines. They have the specific advantage that such ships could remain at sea for long periods of time without needing to refuel. But small modular reactors have never taken off as a source of grid energy. The prevailing opinion for why this is seems to be that they are simply not cost effective. Larger reactors,  which are already expensive endeavors, produce more megawatts per dollar. SMRs are simply too cost inefficient.

This is unfortunate because they have a lot of advantages. Their initial investment is smaller, even though the cost per unit energy is more. They are safe and reliable. They have a small footprint. And they are scalable. The military uses them because the strategic advantages are worth the higher cost. Some argue that the zero carbon on demand energy they provide is worth the higher cost, and I think this is a solid argument. Also there are continued attempts to develop the technology to bring down the cost. Arguably it may be worth subsidizing the SMR industry so that the technology can be developed to greater cost effectiveness. Decarbonizing the energy sector is worth the investment.

But there is another question – are there civilian applications that would also justify the higher cost per unit energy? I have recently encountered two that are interesting. The first is a direct extension of the military use – using an SMR to power a cargo ship. South Korean company, HD Korea Shipbuilding & Offshore Engineering, has revealed their designs for an SMR powered cargo ship, and has received “approval in principle”. Obviously this is just the beginning phase – they need to actually develop the design and get full approval. But the concept is compelling.

Continue Reading »

The flying car is an icon of futuristic technology – in more ways than one. This is partly why I can’t resist a good flying car story. I was recently sent this YouTube video on the Alef flying car. The company says his is a street-legal flying car, with vertical take off and landing. They also demonstrate that they have tested this vehicle in urban environments. They are available now for pre-order (estimated price, $300k). The company claims: “Alef will deliver a safe, affordable vehicle to transform your everyday commute.” The claim sounds reminiscent of claims made for the Segway (which recently went defunct).

The flying car has a long history as a promise of future technology. As a technology buff, nerd, and sci-fi fan, I have been fascinated with them my entire life. I have also seen countless prototype flying cars come and go, an endless progression of overhyped promises that have never delivered. I try not to let this make my cynical – but I am cautious and skeptical. I even wrote an entire book about the foibles of predicting future technology, in which flying cars featured prominently.

So of course I met the claims for the Alef flying car with a fair degree of skepticism – which has proven entirely justified. First I will say that the Alef flying car does appear to function as a car and can fly like a drone. But I immediately noticed in the video that as a car, it does not go terribly fast. You have to do some digging, but I found the technical specs which say that it has a maximum road speed of 25 MPH.  It also claims a road range of 200 miles, and an air range of 110 miles. It is an EV with a gas motor to extend battery life in flight, with eight electric motors and eight propellers. It is also single passenger. It’s basically a drone with a frame shaped like a car with tires and weak motors – a drone that can taxi on roads.

It’s a good illustration of the inherent hurdles to a fully-realized flying car of our dreams, mostly rooted in the laws of physics. But before I go there, as is, can this be a useful vehicle? I suppose, for very specific applications. It is being marketed as a commuter car, which makes sense, as it is single passenger (this is no family car). The limited range also makes it suited to commuting (average daily commutes in the US is around 42 miles).

Continue Reading »

I am fascinated by the technologies that live largely behind the scenes. These are not generally consumer devices, but they may be components of consumer products, or may largely have a role in industry – but they make our modern world possible, or make it much better. In addition I think that material science is largely underrated in terms of popular appeal, but it is material science that often make all other technologies possible or feasible. There is another aspect of technology that I have been increasingly interested in – solid state technology. These are, generally speaking, devices that use electricity rather than moving parts. You are likely familiar with solid state drives, that do not have spinning discs and therefore are smaller, use less power, and last longer. One big advantage of electric vehicles is that they are largely solid state, without the moving parts of an engine.

There is a technology that combines all three of these features – it is a component technology, dependent on material science, and solid state: thermoelectric devices. This may not sound sexy, but bear with me, this is cool (pun intended) technology. Thermoelectric materials are those that convert electricity into a temperature difference across a material, or convert a temperature difference into electricity. In reality, everything is a thermoelectric material, but most materials have insignificant thermoelectric effects (so are functionally not thermoelectric).

Thermoelectric devices can be used to harvest energy, from any temperature difference. These are generally not large amounts of energy – we don’t have thermoelectric power plants connected to the grid – and they are currently not practical and cost effective enough for a large scale. This may be possible in the future, but not today. However, for applications that require small amounts of energy, harvesting that energy from ambient sources like small temperature differences is feasible.

Continue Reading »

My younger self, seeing that title – AI Powered Bionic Arm – would definitely feel as if the future had arrived, and in many ways it has. This is not the bionic arm of the 1970s TV show, however. That level of tech is probably closer to the 2070s than the 1970s. But we are still making impressive advances in brain-machine interface technology and robotics, to the point that we can replace missing limbs with serviceable robotic replacements.

In this video Sarah De Lagarde discusses her experience as the first person with an AI powered bionic arm. This represents a nice advance in this technology, and we are just scratching the surface. Let’s review where we are with this technology and how artificial intelligence can play an important role.

There are different ways to control robotics – you can have preprogrammed movements (with or without sensory feedback), AI can control the movements in real time, you can have a human operator, through some kind of interface including motion capture, or you can use a brain-machine interface of some sort. For robotic prosthetic limbs obviously the user needs to be able to control them in real time, and we want that experience to feel as natural as possible.

The options for robotic prosthetics include direct connection to the brain, which can be from a variety of electrodes. They can be deep brain electrodes, brain surface, scalp surface, or even stents inside the veins of the brain (stentrodes). All have their advantages and disadvantages. Brain surface and deep brain have the best resolution, but they are the most invasive. Scalp surface is the least invasive, but has the lowest resolution. Stentrodes may, for now, be the best compromise, until we develop more biocompatible and durable brain electrodes.

Continue Reading »

On January 20th a Chinese tech company released the free version of their chatbot called DeepSeek. The AI chatbot, by all accounts, is about on par with existing widely available chatbots, like ChatGPT. It does not represent any new abilities or breakthrough in quality. And yet the release shocked the industry causing the tech-heavy stock market Nasdaq to fall 3%. Let’s review why that is, and then I will give some thoughts on what this means for AI in general.

What was apparently innovative about DeepSeek is that, the company claims, it was trained for only $8 million. Meanwhile ChatGPT 4 training cost over $100. The AI tech industry is of the belief that further advances in LLMs (large language models – a type of AI) requires greater investments, with ChatGPT-5 estimated to cost over a billion dollars. Being able to accomplish similar results at a fraction of the cost is a big deal. It may also mean that existing AI companies are overvalued (which is why their stocks tumbled).

Further, the company that made DeepSeek used mainly lower power graphics chips. Apparently they did have a horde of high end chips (the export of which are banned to China) but was able to combine them with more basic graphics chips to create DeepSeek. Again, this is what is disruptive – they are able to get similar results with lower cost components and cheaper training. Finally, this innovation represents a change for the balance of AI tech between the US and China. Up until now China has mainly been following the US, copying its technology and trailing by a couple of years. But now a Chinese company has innovated something new, not just copied US technology. This is what has China hawks freaking out. (Mr. President, we cannot allow an AI gap!)

Continue Reading »

My recent article on social media has fostered good social media engagement, so I thought I would follow up with a discussion of the most urgent question regarding social media – should the US ban TikTok? The Biden administration signs into law legislation that would ban the social media app TikTok on January 19th (deliberately the day before Trump takes office) unless it is sold off to a company that is not, as it is believed, beholden to the Chinese government. The law states it must be divested from ByteDance, which is the Chinese parent company who owns TikTok. This raises a few questions – is this constitutional, are the reasons for it legitimate, how will it work, and will it work?

A federal appeals court ruled that the ban is constitutional and can take place, and that decision is now before the Supreme Court. We will know soon how they rule, but indicators are they are leaning towards allowing the law to take effect. Trump, who previously tried to ban TikTok himself, now supports allowing the app and his lawyers have argued that  he should be allowed to solve the issue. He apparently does not have any compelling legal argument for this. In any case, we will hear the Supreme Court’s decision soon.

If the ban is allowed to take place, how will it work? First, if you are not aware, TikTok is a short form video sharing app. I have been using it extensively over the past couple of years, along with most of the other popular platforms, to share skeptical videos and have had good engagement. Apparently TikTok is popular because it has a good algorithm that people like. TikTok is already banned on devices owned by Federal employees. The new ban will force app stores in the US to remove the TikTok app and now allow any further updates or support. Existing TikTok users will continue to be able to use their existing apps, but they will not be able to get updates so they will eventually become unusable.

Continue Reading »

One of the things I have come to understand from following technology news for decades is that perhaps the most important breakthroughs, and often the least appreciated, are those in material science. We can get better at engineering and making stuff out of the materials we have, but new materials with superior properties change the game. They make new stuff possible and feasible. There are many futuristic technologies that are simply not possible, just waiting on the back burning for enough breakthroughs in material science to make them feasible. Recently, for example, I wrote about fusion reactors. Is the addition of high temperature superconducting material sufficient to get us over the finish line of commercial fusion, or are more material breakthroughs required?

One area where material properties are becoming a limiting factor is electronics, and specifically computer technology. As we make smaller and smaller computer chips, we are running into the limits of materials like copper to efficiently conduct electrons.  Further advance is therefore not just about better technology, but better materials. Also, the potential gain is not just about making computers smaller. It is also about making them more energy efficient by reducing losses to heat when processors work. Efficiency is arguably now a more important factor, as we are straining our energy grids with new data centers to run all those AI and cryptocurrency programs.

This is why a new study detailing a new nanoconducting material is actually more exciting than it might at first sound. Here is the editor’s summary:

Noncrystalline semimetal niobium phosphide has greater surface conductance as nanometer-scale films than the bulk material and could enable applications in nanoscale electronics. Khan et al. grew noncrystalline thin films of niobium phosphide—a material that is a topological semimetal as a crystalline material—as nanocrystals in an amorphous matrix. For films with 1.5-nanometer thickness, this material was more than twice as conductive as copper. —Phil Szuromi

Continue Reading »