Shelly Palmer’s Blog

ShellyPalmer

THINK ABOUT THIS

Shelly Palmer

This is the time of year that everyone makes lofty sci-fi predictions about the future: flying cars, practical AI, nuclear fusion… hold on a sec… these are no longer lofty sci-fi predictions. There are at least a dozen passenger drones (flying cars) waiting for FAA approval. ChatGPT, Midjourney, Stable Diffusion, and dozens of other practical AI solutions have become available in the past year. Just a few weeks ago, scientists at the Lawrence Livermore National Laboratory announced the first-ever demonstration of fusion “ignition,” a process where more energy was generated than was needed to trigger the reaction.

The news is filled with negative stories, and doom-scrolling seems to have become a national pastime, but my predictions for 2023 begin with an optimistic belief that technology will continue to improve at an exponential rate: continuous improvement that will not only lead to greater productivity, but offer us greater choices about how we want to spend our days.

Personal drones may not become the flying cars as imagined in The Jetsons, but they will be used for rescue missions and reduce the time needed to accomplish a variety of transportation tasks.

AGI (artificial general intelligence), which is supposed to describe AI models that have an ability to learn that is equal to a human, may not happen as imagined in the movies, but AI is proving itself to be an immense productivity enhancer and a wonderful creative toolbox.

While limitless fusion energy may be years (or even decades) away, it is possible. Practical fusion will end our dependence on fossil fuels and make fresh water available for everyone and everything. Commercialized fusion energy will usher in a new era as it changes how we do absolutely everything.

I’m still finalising my list of predictions for 2023 and beyond. In the meantime, I’d welcome the opportunity to hear your ideas about our near-term future.

Neuroscientists have created a mood decoder that can measure depression

By guest author Jessica Hamzelou from MIT Tech

Courtesy of Dr Sameer Sheth/Baylor College of Medicine

John’s life changed forever when he broke up with his girlfriend. The breakup sent him into a downward spiral, and led to his first depressive episode when he was 27 years old. “At first it’s just extreme sadness … then you start losing sleep,” says John (not his real name), who spoke on condition of anonymity. He developed crippling anxiety and experienced panic attacks and dark thoughts that eventually led him to attempt to end his own life.

Drugs didn’t work for John—he says he has tried pretty much every antidepressant, antipsychotic, and sedative out there. And while electroconvulsive therapy—a treatment that delivers electrical stimulation to one or both sides of a person’s head—eventually pulled him out of his first depressive episode, it didn’t touch the symptoms of his second episode, which started around five years later.

But as part of a clinical trial, John has benefited from an experimental treatment that involves inserting electrodes deep into his brain to deliver regular pulses of electricity. Deep brain stimulation is already used to treat severe cases of epilepsy and a few movement disorders such as Parkinson’s. But depression is more complicated—partly because we still don’t fully understand what’s going on in the brain when it occurs.

“Depression is a complex illness,” says Patricio Riva Posse, a neurologist at the Emory School of Medicine in Atlanta, Georgia, who was not involved in the trial. “It’s not like trying to correct one tremor—there’s a whole universe of symptoms.” These include low mood, suicidality, inability to experience pleasure, and changes in motivation, sleep, and appetite.

Doctors have been using electricity to treat brain disorders—including depression—for decades, and some studies have found that electrodes placed deep inside the brain can jolt some people out of their symptoms. But results vary. Neuroscientists hope that by getting a better idea of what’s happening inside the brains of people with symptoms like John’s, they can make the treatment more effective.

John is one of five people who have volunteered to have their brains probed as part of a clinical trial. At the start of 2020, he had a total of 14 electrodes implanted across his brain. For nine days, he stayed in a hospital with protruding cables wrapped around his head, while neuroscientists monitored how his brain activity correlated with his mood.

The researchers behind the trial say they have developed a “mood decoder”—a way of being able to work out how someone is feeling just by looking at brain activity. Using the decoder, the scientists hope to be able to measure how severe a person’s depression is, and target more precisely where the electrodes are placed to optimize the effect on the patient’s mood. So far, they have analyzed the results of three volunteers.

What they have found is extremely promising, says Sameer Sheth, a neurosurgeon based at Baylor College of Medicine in Houston, Texas, who is leading the trial. Not only have he and his colleagues been able to link volunteers’ specific brain activity with their mood, but they have also found a way to stimulate a positive mood. “This is the first demonstration of successful and consistent mood decoding of humans in these brain regions,” says Sheth. His colleague Jiayang Xiao presented the findings at the Society for Neuroscience’s annual meeting in San Diego in November.

Zapping depression

Deep brain stimulation (DBS) usually involves placing one or two electrodes deep into the brain to deliver pulses of electricity to specific regions. It can work really well for some people with Parkinson’s disease, where it’s used to stimulate areas that control movement. Researchers are exploring whether it might also help treat psychiatric issues including obsessive-compulsive disorder, eating disorders, and depression.

A handful of studies performed in the early and mid 2000s suggested that DBS could help people with depression that didn’t respond to typical treatments, like antidepressants. But initial results of two large clinical trials were disappointing, and the tests were stopped early.

t’s not clear why these trials didn’t see the same results as earlier studies. But the varying success rates might have something to do with how the brain stimulation is delivered. Several brain regions are thought to play a role in depression. And there are lots of potential ways to deliver electrical pulses. “We don’t know how to deliver DBS intelligently to any given individual [with depression],” says Sheth. “This is just a very immature therapy.”

Sheth has been trying to figure out what might work best. He and his colleagues have borrowed a brain surgery approach that is sometimes used to help people with epilepsy who don’t get better with drugs.

In these cases, doctors might implant electrodes across the person’s brain in order to find out where the seizures are starting. Once identified, these regions can either be stimulated with electrodes or removed entirely.

Depression doesn’t originate from a single point in the brain, the way a seizure does. But Sheth and his colleagues are taking the same approach—temporarily implanting electrodes across the brain to monitor brain activity—for insight into the condition.

The team is particularly interested in how patterns of brain activity differ when a person is feeling better or especially low. Sheth and his colleagues are also experimenting with stimulation—what level, type, and frequency works best to get the brain back to a positive mood state? Armed with this information, neurosurgeons will be in a much better position to help people with depression, and deep brain stimulation is more likely to work, says Sheth.

Back online

John was the first trial volunteer to undergo the procedure. Sheth and his colleagues put him under general anesthetic before drilling holes into his skull to insert the electrodes. The team implanted two DBS electrodes on each side of the brain in regions thought to be involved in symptoms of depression. An additional five temporary electrodes were placed on each side of the brain to monitor John’s activity in regions linked to mood and cognition.

To find the right place to stimulate, the team needed to wake John up during his operation. He remembers being repeatedly asked how he felt as surgeons probed his brain with electrodes. “Then they hit a spot and I said: ‘I actually feel back online,’” he says. “Depression is like a constant weight on your soul. When they touched that perfect little spot, that weight lifted.”

He remembers hearing the doctors laugh and tell him they’d found the right place, and then falling back asleep.

John woke up “with a headache like nothing ever before” and spent the next nine days being closely monitored by Sheth and his colleagues. Every few hours, the medical team would ask him questions about his mood and how he was feeling.

At the end of the nine days, the team removed the 10 monitoring electrodes from John’s brain but left in the four DBS electrodes. These electrodes were connected to a rechargeable battery implanted in John’s chest. In the years since, the pulses of stimulation have been tweaked slightly. Six months after the operation, the team turned off the stimulation without telling John. His symptoms immediately worsened. “It was obvious,” he says. “I told them: ‘I don’t know what you did, but I can’t sleep, I’m anxious … it’s not working.”

The device was switched back on and has been running ever since. Sheth’s team has carried out the same procedure in four other people so far—all of them with severe, treatment-resistant depression. They plan to study 12 people in total.

Despite the early signs of success, Sheth and his colleagues don’t plan to carry out this same procedure more widely. Temporarily implanting 10 electrodes into the brain provides insight into a specific person’s depression, but it isn’t a practical approach for a condition that affects almost 3 million people in the US alone. It is an invasive, expensive procedure that takes a lot of time and carries risks.

Instead, Sheth hopes to find trends among his 12 volunteers and use these to develop an improved form of DBS that can help anyone who needs it. “We’re hoping that there are some generalizable findings that we get out of this,” he says.

Sheth and his colleagues have analyzed brain recordings from only three people so far, but they are already seeing trends. A brain region called the cingulate cortex fires in a certain way when all three are in a better mood and shows the opposite pattern of activity when the volunteers are experiencing a low mood, says Sheth.

Riva Posse says that the results are “very encouraging.” We are starting to understand that “there are signals of depression that seem to be pretty consistent across the board,” he says. “This is going to advance, considerably, the understanding of depression and help come up with … neurostimulation approaches.”

Still, it is still too soon to say whether these findings will track across a larger group of people. “It’s only three patients,” says Darin Dougherty, a psychiatrist at Mass General Research Institute in Boston who specializes in neurosurgery for depression.

Dougherty thinks that Sheth’s research is “essential.” He adds, “Hopefully they can get enough data from a small group of people so that we can move away from [implanting multiple temporary electrodes].” He predicts that Sheth’s approach might identify a brain region that will be worth targeting in most people with treatment-resistant depression, and that noninvasive brain scans will find the exact spot to implant the electrode.

Measuring mood

Sheth and his colleagues also found some differences among the three volunteers, and the team’s “mood decoder” could identify how each volunteer was feeling based on their brain activity.

He hopes that in the future, new technologies will allow him and others to collect this information noninvasively, perhaps using a device that sits on a person’s head. Such a device could be used to measure the severity of a person’s symptoms, he says.

Today, a person with symptoms of depression will typically be asked a series of questions to determine the severity of the condition. Having some kind of objective measure—such as readings from a brain scan—is a key goal for psychiatry, says Dougherty.

It could also be problematic, though. Brain scans might never be sensitive enough to account for individual differences in people’s brains when it comes to symptoms of depression, and they might miss signs in some people and overestimate them in others. Sheth also acknowledges the possibility that because of research like his, brain scans could one day be used to diagnose depression in someone who is not obviously unwell or reveal it in someone who doesn’t want it known.

John, for one, doesn’t want others—particularly potential employers—to know he has a history of depression. “People don’t understand depression, and unfortunately, they see it as a weakness,” he says.

“You can’t really argue that… we should not try to help all these millions and millions of people with depression… just because there’s a possibility of misuse,” says Sheth. “We have to find ways of helping these folks. The rest of society can help us put guardrails on how this technology should be used.”

John’s electrodes are still delivering pulses of electrical stimulation deep in his brain. He charges the battery embedded in his chest every week. “As far as I know, if the stimulation stops, I go back to square one,” he says. And while DBS might not work for everyone for depression, “it saved my life,” he says.

 

www.technologyreview.com

Shelly’s Blog: How Far Can AI Go?

 

December 18, 2022 by Shelly Palmer

If you haven’t already uploaded your 10-to-20 selfies to Lensa and let it turn you into a superhero, or a rock star, or an astronaut, you should give it a try. It’s fun, but it’s also instructive. You’ll learn about the type and quality of inputs generative AI needs in order to obtain satisfactory results, and you’ll experience the workflow and process of consumer-grade generative AI. But most importantly, when the model transforms you into a mystical creature in a cosmic setting, or an anime character, or a cyborg, you will find yourself asking one question: How far can this technology go?

Tweet Article Share Article on Facebook Email Article

Share Article on LinkedIn

If you haven’t already uploaded your 10-to-20 selfies to Lensa and let it turn you into a superhero, or a rock star, or an astronaut, you should give it a try. It’s fun, but it’s also instructive. You’ll learn about the type and quality of inputs generative AI needs in order to obtain satisfactory results, and you’ll experience the workflow and process of consumer-grade generative AI. But most importantly, when the model transforms you into a mystical creature in a cosmic setting, or an anime character, or a cyborg, you will find yourself asking one question: How far can this technology go?

A Quick AI Tool Check

Let’s say we wanted to create, produce, and distribute a video. Where might we use AI to assist us?

Scriptwriting: AI can be used to generate new storylines or suggest changes to existing ones, using natural language processing and machine learning algorithms.

Content creation: AI can be used to generate visual effects, such as background landscapes or special effects, or to create realistic 3D models of objects and characters.

Audio and video editing: AI can be used to automate tasks such as transcribing dialog, synchronizing audio and video, and identifying and removing background noise.

Sound mixing: AI can be used to help mix the final audio balancing dialog, sound effects, and music.

Audio processing: AI can be used to enhance the audio quality. It can also be used to help producers mimic the audio styles and sonic landscapes of existing masterworks. And it can be used to create new, never-heard-before sonic landscapes.

Language translation: AI can be used to translate content into different languages, using machine translation algorithms.

Quality control: AI can be used to automatically identify errors or inconsistencies in the production process, such as frame rates, dropout, lighting or sound issues.

Marketing and promotion: AI can be used to analyze viewer data and create personalized recommendations for shows, or to target ads to specific demographics.

SEO/SEM: AI can be used to generate key words and phrases and generally improve all aspects of search engine optimization and marketing.

Content recommendation: AI can be used to recommend content to viewers based on their preferences and viewing history, using techniques such as collaborative filtering and recommendation algorithms.

Audience analysis: AI can be used to analyze audience data and provide insights on viewer behavior, using techniques such as natural language processing and machine learning.

Automated content moderation: AI can be used to automatically flag inappropriate or offensive content, using techniques such as image and video analysis.

Virtual assistants: AI can be used for a clerical work such as resource and meeting scheduling as well as other productivity enhancements.

Note: This list is remarkably incomplete — in practice, there are already dozens of AI models at work for us — we just don’t interface with them directly.

What’s Next?

All of the areas mentioned above are the epicentres of nascent AI-assisted ecosystems. But at the moment, most of the available technology comes in the form of purpose-built, narrowly focused tools. For example: Adobe has a suite of AI tools you can access directly inside of Photoshop. Each tool does something specific (such as removing a background or changing the sky).

The roadmap is pretty obvious. Each purpose-built AI tool will improve at an exponential rate. Over time, we’ll start to see interfaces that allow us to layer and script various disparate AI tools. You can think of it as an AI model that is trained to understand the problem you are trying to solve, then sends data to other AI models and organises the various outputs into results you can use. More simply stated, we can look forward to AI models that will harness the power of other AI models which will in turn harness the power of other AI models until we get to a point where AI-assisted productivity will be table stakes.

What happens then?

If you look back in 30-year blocks and think about what technology was available in 1990, 1960, 1930, 1900, 1870, and 1840, you’ll get a sense of how things may change. I don’t know any more about the future than anyone else, but I do know that today, you are experiencing the slowest rate of technological change you will ever experience for the rest of your life. So take a few moments to jump into the AI tools that are directly related to the work you are doing right now. Once you start to engage with AI, you’ll be amased at where your imagination takes you.

Author’s note:This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. I am not a financial advisor. Nothing contained herein should be considered financial advice. If you are considering any type of investment you should conduct your own research and, if necessary, the advice of a licensed financial advisor.

www.shellypalmer.com