Dear Aventine Readers,

The astonishing — and frightening — capabilities of new generative AI systems like ChatGPT made artificial intelligence the story of the year, at least as far as technology goes. Much of the world was either using it, worrying about it, investing in it, writing about it or pushing to regulate it — sometimes all at once! But 2023 was remarkable as well for many other stunning advances. To wrap up the year we’ve chosen what we believe are the twelve most significant breakthroughs in science and technology of the past twelve months, all of which could have long-term implications for the way we all live in the years to come.

Thanks for reading, and have a happy new year!

Sincerely,

Danielle Mattoon
Executive Director, Aventine

Subscribe

Subscribe to our newsletter and be kept up to date on upcoming Aventine projects

If there has been a defining story in technology this year, it is the democratization of artificial intelligence. In the year-plus since OpenAI’s ChatGPT was made public, it and other large language models — built by companies like Google, Meta and Anthropic — have “changed the landscape in ways that I don't think people really expected,” said David Ferrucci, who led the team that built IBM’s Watson, the AI system that beat Ken Jennings in “Jeopardy!” in 2011, and is now the CEO of an AI startup. 

Powerful and user friendly, the models have become popular with everyone from high school students to Fortune 500 CEOs. And though there are plenty of concerns about the widespread use of these new tools, as well as a reasonable consensus that LLMs have not achieved human-level intelligence, many experts agree that the current ubiquity of these tools means that AI has become integral to how the world works. “AI has become a sort of essential part of our socioeconomic and personal lives. No one expects a future where AI doesn't play a role, in every aspect — your job, the economy, your relationships, your personal life. It's, it's going to be a pervasive thing.” said Ferrucci. ”[AI has become] the transformative force that will shape the future.”

For years, brain-computer interfaces, which allow paralyzed people to move and communicate, have been employed selectively with astonishing results. This year they became significantly more effective and user friendly, paving the way for wider use. In May, researchers at the University of Lausanne in Switzerland announced that they were able to help a man whose legs had been paralyzed to walk again by creating a new connection between his brain and spinal cord. This was previously possible only through brain or spinal-cord stimulation, so the new approach — as well as being a technological milestone — is notable for providing the man with personal control over his movements.

Meanwhile, a team of researchers across multiple U.S. institutions created a brain implant that helps people who cannot otherwise communicate verbalize through text-to-speech at rates that are finally useful, according to research published in September in Nature. Previous approaches allowed users to verbalize 18 words per minute; the new one allows for speeds as high as 78. And possibly most significant: In May a team at the University of Texas at Austin was able to detect and decode language through noninvasive brain-scanning tools. Don’t call it a mind-reading system just yet, but it’s a major advance over the current state of the art for reading thoughts, which requires invasive surgery to place sensors on the brain. While noninvasive portable brain scanners are far from commonplace or accessible, researchers in the field see this year’s breakthrough as an indicator that BCIs could become increasingly useful, effective and powerful in the coming years if they receive the necessary investment and oversight.

CRISPR has been heralded as perhaps the most important biotechnological breakthrough of the last decade, providing a relatively simple, affordable yet precise means of editing genes. The technology has wide chemical and medical applications, like editing the genes of plants and animals, but the true promise of CRISPR has always been that it could be used to create new therapies for diseases that are currently difficult, expensive or otherwise impossible to treat. This year, for the first time, regulators in both the U.K. and the U.S. have approved a gene therapy for sickle cell disease that relies on CRISPR technology. 

Sickle cell is an inherited disease that causes misshapen red blood cells, which in turn usually leads to anemia, periods of extreme pain, and fatigue. Current treatment for the disease may include regular blood transfusions that are not entirely effective at reducing symptoms. The new CRISPR-based therapy promises a reduction in the crippling pain of sickle cell disease for at least a year, according to interim trial results and reports from patients. It is able to reactivate the gene for fetal hemoglobin production that is normally suppressed in adults, thus allowing the body to produce fetal hemoglobin undamaged by sickle cell disease. 

The new therapy is not without challenges, however. It could cost millions of dollars per patient, requires staff with advanced medical training to administer it and is onerous for patients, as it involves intense chemotherapy. One concern for any CRISPR therapy is that it could create random genetic mutations or edits beyond the intended target, and researchers are still observing early recipients — the first patient received the drug in 2019 — to determine whether that particular concern will bear out here. Still, the approval of the therapy could herald the beginning of a new wave of CRISPR-based drug development, allowing us to tackle diseases that have until now remained almost impossible to treat.

Beginning on Halloween this past year, Portugal set a record for the longest continuous length of time the country produced enough renewable energy to meet all of its electricity demands — 149 straight hours, or a little over six days. Perhaps even more impressive is that for 95 of those hours, it made more energy than it needed, exporting the excess to neighboring Spain. 

While other countries, including Albania and Iceland, have been powering their electrical needs with nearly 100 percent renewable sources for years, many of those nations mostly rely on one dominant energy source, usually hydroelectric or hydrothermal. Portugal’s energy portfolio is more diverse, making the achievement an important model for other nations attempting to meet Paris 2050 targets. 

Portugal got started on its goal of reaching net-zero emissions by 2050 back in 2016, several years before most other European countries, which gave it a head start on investments and build-out. (The October achievement was, to be fair, partly due to timing, as energy demands tend to be lower in the fall and spring.) And though surviving for six days on renewable energy alone is nowhere near what will be required in a few decades, the achievement proves that Portugal’s diverse mix of renewable power sources and its continued investment are now showing results, providing an example of what’s possible for net-zero power generation.

Concerns about how AI should be regulated aren’t new, but in 2023 they gained traction. “ChatGPT changed the conversation,” explained Ann Kristin Glenster, Senior Policy Advisor on Technology Governance and Law at the University of Cambridge’s Minderoo Centre for Technology and Democracy. As soon as ordinary people could experience AI firsthand, formerly niche concerns about the technology’s ability to invent facts, spread disinformation, exaggerate bias and replace workers became mainstream issues, prompting lawmakers around the globe to deliver proposals aimed at mitigating AI’s potential harms. Though such efforts can seem initially toothless, Glenster said that they tend to “follow on the back end of much harder regulatory policies that are being debated in the backroom.” 

A clearer picture of those sorts of conversations is now emerging in Europe, where lawmakers are in the final stretch of drawing up the European Union’s AI Act, which was provisionally passed in December 2023 and is expected to phase into effect over the next two years. It is, said Glenster, who has studied the legislation, far more rigorous than anything currently being developed elsewhere, banning certain uses of AI and putting in place specific rules for so-called high-risk applications of the technology. Until now, AI development and deployment have gone largely unconstrained; pretty soon, at least in Europe, there will be regulations to ensure that any product making use of AI does so safely, fairly and transparently — requirements, according to Glenster, that many companies will undoubtedly find burdensome.

Human brain cells are a mystery compared to cells in the rest of the body. We know very little about how they function, when and why they respond to stimuli and even what they’re made of and how they’re shaped. This year, teams totaling hundreds of researchers from across the world took a major step in tackling some of these unknowns by mapping and identifying more than 3,000 types of brain cells that were never before understood, and sharing the results in the largest atlas of the human brain ever published. 

The atlas identifies the location and sometimes the structure and function of specific cells, as well as how gene expression changes in certain cells compared to similar cells found in primates or mice. Despite the size of the undertaking, many brain cells remain unidentified or are so far not linked to their potential function. One hope for the research is that it will provide greater understanding of how human and animal brains are different, which would illuminate the ways in which treatments for one may or may not work for the other. Eventually, new understanding of degenerative brain diseases and new ways to treat them might emerge.

When an Indian spacecraft gently tapped a lander down onto the surface of the moon in August, it made India the fourth country ever to lay claim to such a feat — following the Soviet Union, the United States and China. But more important, it became the first country to put a lander near the moon’s south pole, a completely unexplored area believed to be critical to future human exploration efforts because it might hold frozen water deposits. It also happens to be so heavily cratered that landings are difficult, which is why this landing was particularly notable. India aims to send people to the moon by 2040, making it part of a new age of moon exploration, with more countries competing for bragging rights in space. Planned human moon landings from the U.S. and China are scheduled for the next two decades; following in India’s footsteps, both are reportedly eyeing landing areas near the south pole.

 Artificial intelligence is increasingly making its way into military hardware, turning what once seemed like a problem for another generation into an ethical minefield, as well as the topic of important United Nations conversations around curbing it. The core of the issue is whether — or at least, to what extent — humans should allow weapons systems to make life-and-death decisions autonomously. It’s increasingly likely that, say, a gunsight could provide a sniper with automated target identification; not far behind may be a drone swarm that is deployed to autonomously seek and destroy a target. 

Now, as The New York Times reported in November, the United Nations is attempting to find a path toward legally binding rules that will dictate how militaries can use these technologies. There are two large issues here, though. First, as MIT Technology explored earlier this year, the ethical nuances around the technology, beyond binary “for” and “against” positions, are deeply complex, so that it is hard to find alignment. Second, national defense priorities vary hugely around the globe, making it even harder to reach a consensus. What is increasingly clear, though, as AI continues to mature at a rapid pace, is that the AI-powered warfare is becoming far more plausible, giving the U.N.’s efforts more urgency.

Just over 80 miles off the east coast of England, there’s a giant sandbank. Dogger Bank, as it’s known — named for the fishing ships that trawled its waters — has become home to the world’s largest wind farm. Just 60 feet to 207 feet below the ocean’s surface, the site offers a rare combination of shallow waters and steady, powerful winds, making it an ideal site for wind power. Each rotation of the turbines’ three 351-foot blades, according to Equinor, the lead operator of the wind farm, generates enough energy to power an average British home for two days. When it is fully operational in 2026, the wind farm will include 277 turbines and be capable of generating 3.6 gigawatts of power. Given that the license to develop the wind farm was granted in 2010, many glasses were raised when the first turbine became operational on October 7, sending power to the UK’s national grid.

The chance to hail a flying taxi got a lot better this year. Companies around the world have been busy developing electric vertical takeoff and landing (eVTOL) vehicles in recent years, and 2023 saw several manufacturers — including Santa Clara-based Archer and Vermont-based Beta Technologies — perform initial test flights of full-scale prototypes or pre-production models. Most of these aircraft resemble a cross between a Cessna lightweight airplane and a quadcopter drone, with a bubble-shaped passenger compartment flanked by propellers that allow for vertical takeoff and horizontal flight. Perhaps more notable than the test flights, though, is the increasing willingness among regulators to certify these aircraft. 

Leading the charge in that respect is China, whose Civil Aviation Administration this fall granted a company called EHang permission to test the world’s first eVTOL taxi, which also happens to be fully autonomous. To reach this stage, according to The Economist, EHang has completed 40,000 test flights and provided numerous safety reassurances, including data from crash tests. The company said that it hopes to offer sightseeing trips around Guangdong by the end of this year. Meanwhile, the Federal Aviation Administration has laid out plans to allow small-scale trials of nonautonomous versions in the U.S. as soon as 2025, envisioning competitive commercial operations in at least one location in the country by 2028. These vehicles may seem like a frivolity for now — and might for some time — but it’s undeniably the case that the hardware and accompanying regulation are finally catching up with enthusiasm for eVTOL technology.

For decades, chemists tasked with building complex molecules, such as those that make up pharmaceutical drugs, have had to develop complicated step-by-step chemical reactions to assemble atoms into the necessary structures, which can take weeks or months or even prove impossible. Every time work begins on the creation of a new drug, the painstaking process must start from scratch. That has become increasingly problematic as computational approaches have become the norm for designing new compounds: A theoretical chemist can design a fantastically useful molecule using computational techniques only to be told by an applied chemist that it would be prohibitively time-consuming and expensive to produce. 

Now, a shortcut is on the horizon. An emerging field of chemistry called skeletal editing is allowing scientists to test out new molecular structures by adding, removing or swapping individual atoms buried inside already existing structures. It’s still early days for the technique: It doesn’t work for all molecules and so far can perform only specific kinds of edits, so early critics are eager to point out that the technique could ultimately be of limited use. Its proponents, however, believe that it could allow chemists to develop previously unbuildable molecules and speed up the creation of new drugs.

Those of us who feel chained to our smartphones got a glimpse of an alternative future in November from a company called Humane. The company’s Ai Pin, developed by a pair of former Apple employees, is an experiment in how we could replace the smartphone with something more intimate and less cumbersome. Controlled by voice, touch or by interacting with its laser-projected display, the Pin makes use of the rapidly increasing sophistication of artificial intelligence to understand what a user wants to do and help get it done. 

For now, it can replicate many of the functions of a smartphone: send text messages, make calls, play songs, take photos. Some proposed features, such as object recognition — which would be a more compelling differentiator than the phone options — don’t work yet, but are promised to arrive soon. Costing at least $699 along with a $24 monthly subscription, it’s not exactly an affordable taste of the future. Nor is it a sure hit: Given the high failure rate of new consumer technologies, its success is far from guaranteed. But rather than focusing on the Ai Pin in and of itself, you can think of the device as an alluring peek at how we might interact with AI-powered and AI-inspired personal technology in the future.