Category: Science & Tech

When Cole first received phage therapy, she had been dealing with a blood infection for nearly a month. Her doctors tried a variety of antibiotics with no effect. But 24 hours after they administered phage therapy, Cole’s infection was gone. She seemed cured.

About a month later, however, the infection returned. So the researchers found another phage that would work against the Enterococcus bacteria causing Cole’s infection, and began administering both phages. That seemed to do the trick.

For four months, Cole was infection-free. She left the hospital and went on vacation with her family. But then the infection returned. Cole was out of options. She entered hospice, and seven months later she died of pneumonia.

Van Tyne and her colleagues have spent the past couple of years trying to explain why their phages failed. They don’t yet have an answer, but they do have a hypothesis. A couple of weeks after Cole began receiving the second phage, she developed antibodies against both phages. “Possibly that played a role in limiting how well they were able to find their bacterial targets and kill them,” says Madison Stellfox, a physician and postdoc in Van Tyne’s lab. She posits that perhaps the antibodies coated the phages so they couldn’t enter the bacteria. Or maybe they helped the body clear the phages faster, so they didn’t have time to work.

Cole isn’t the only patient Van Tyne and her colleagues at the University of Pittsburgh have treated. Since Van Tyne started her own lab in 2018, she has developed a library that contains about 200 phages, most isolated from Pittsburgh’s wastewater. Those phages target six or seven species of bacteria. They use that library to develop personalized therapies for patients with life-threatening infections. “We’re trying to match clinical isolates from infected patients with phages that are active on them,” Van Tyne says. 

The team has treated nearly 20 patients. Some have cleared their infections. Some, like Cole, have experienced temporary improvements. Some have had no response at all. But reassuringly, no one has been harmed by the therapy itself.  

All these patients were treated under the FDA’s “compassionate use” program, which provides access to investigational therapies for people with life-threatening illnesses. Case studies can provide valuable insights, but they’re not a pathway to regulatory approval. To move phages into mainstream medicine, we need clinical trials.

Alexander Sulakvelidze, president and chief executive officer at the phage company Intralytix, has been working to develop phage products since the 1990s. In the Republic of Georgia, where he was born, phage therapy is routinely used to treat infections.  

It may be some time before we find out. OpenAI’s announcement of Sora today is a tech tease, and the company says it has no current plans to release it to the public. Instead, OpenAI will today begin sharing the model with third-party safety testers for the first time.

In particular, the firm is worried about the potential misuses of fake but photorealistic video. “We’re being careful about deployment here and making sure we have all our bases covered before we put this in the hands of the general public,” says Aditya Ramesh, a scientist at OpenAI, who created the firm’s text-to-image model DALL-E.

But OpenAI is eyeing a product launch sometime in the future. As well as safety testers, the company is also sharing the model with a select group of video makers and artists to get feedback on how to make Sora as useful as possible to creative professionals. “The other goal is to show everyone what is on the horizon, to give a preview of what these models will be capable of,” says Ramesh.

To build Sora, the team adapted the tech behind DALL-E 3, the latest version of OpenAI’s flagship text-to-image model. Like most text-to-image models, DALL-E 3 uses what’s known as a diffusion model. These are trained to turn a fuzz of random pixels into a picture.

Sora takes this approach and applies it to videos rather than still images. But the researchers also added another technique to the mix. Unlike DALL-E or most other generative video models, Sora combines its diffusion model with a type of neural network called a transformer.

Transformers are great at processing long sequences of data, like words. That has made them the special sauce inside large language models like OpenAI’s GPT-4 and Google DeepMind’s Gemini. But videos are not made of words. Instead, the researchers had to find a way to cut videos into chunks that could be treated as if they were. The approach they came up with was to dice videos up across both space and time. “It’s like if you were to have a stack of all the video frames and you cut little cubes from it,” says Brooks.

The transformer inside Sora can then process these chunks of video data in much the same way that the transformer inside a large language model processes words in a block of text. The researchers say that this let them train Sora on many more types of video than other text-to-video models, including different resolutions, durations, aspect ratio, and orientation. “It really helps the model,” says Brooks. “That is something that we’re not aware of any existing work on.”

OpenAI is well aware of the risks that come with a generative video model. We are already seeing the large-scale misuse of deepfake images. Photorealistic video takes this to another level.

Technology use often goes wrong, Parsons notes, “because we’re too focused on either our own ideas of what good looks like or on one particular audience as opposed to a broader audience.” That may look like an app developer building only for an imagined customer who shares his geography, education, and affluence, or a product team that doesn’t consider what damage a malicious actor could wreak in their ecosystem. “We think people are going to use my product the way I intend them to use my product, to solve the problem I intend for them to solve in the way I intend for them to solve it,” says Parsons. “But that’s not what happens when things get out in the real world.”

AI, of course, poses some distinct social and ethical challenges. Some of the technology’s unique challenges are inherent in the way that AI works: its statistical rather than deterministic nature, its identification and perpetuation of patterns from past data (thus reinforcing existing biases), and its lack of awareness about what it doesn’t know (resulting in hallucinations). And some of its challenges stem from what AI’s creators and users themselves don’t know: the unexamined bodies of data underlying AI models, the limited explainability of AI outputs, and the technology’s ability to deceive users into treating it as a reasoning human intelligence.

Parsons believes, however, that AI has not changed responsible tech so much as it has brought some of its problems into a new focus. Concepts of intellectual property, for example, date back hundreds of years, but the rise of large language models (LLMs) has posed new questions about what constitutes fair use when a machine can be trained to emulate a writer’s voice or an artist’s style. “It’s not responsible tech if you’re violating somebody’s intellectual property, but thinking about that was a whole lot more straightforward before we had LLMs,” she says.

The principles developed over many decades of responsible technology work still remain relevant during this transition. Transparency, privacy and security, thoughtful regulation, attention to societal and environmental impacts, and enabling wider participation via diversity and accessibility initiatives remain the keys to making technology work toward human good.

MIT Technology Review Insights’ 2023 report with Thoughtworks, “The state of responsible technology,” found that executives are taking these considerations seriously. Seventy-three percent of business leaders surveyed, for example, agreed that responsible technology use will come to be as important as business and financial considerations when making technology decisions. 

This AI moment, however, may represent a unique opportunity to overcome barriers that have previously stalled responsible technology work. Lack of senior management awareness (cited by 52% of those surveyed as a top barrier to adopting responsible practices) is certainly less of a concern today: savvy executives are quickly becoming fluent in this new technology and are continually reminded of its potential consequences, failures, and societal harms.

The other top barriers cited were organizational resistance to change (46%) and internal competing priorities (46%). Organizations that have realigned themselves behind a clear AI strategy, and who understand its industry-altering potential, may be able to overcome this inertia and indecision as well. At this singular moment of disruption, when AI provides both the tools and motivation to redesign many of the ways in which we work and live, we can fold responsible technology principles into that transition—if we choose to.

For her part, Parsons is deeply optimistic about humans’ ability to harness AI for good, and to work around its limitations with common-sense guidelines and well-designed processes with human guardrails. “As technologists, we just get so focused on the problem we’re trying to solve and how we’re trying to solve it,” she says. “And all responsible tech is really about is lifting your head up, and looking around, and seeing who else might be in the world with me.”

To read more about Thoughtworks’ analysis and recommendations on responsible technology, visit its Looking Glass 2024.

This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff.

“In one way it operates much like our brain does, where not the whole brain activates all the time,” says Oriol Vinyals, a deep learning team lead at DeepMind. This compartmentalizing saves the AI computing power and can generate responses faster.

“That kind of fluidity going back and forth across different modalities, and using that to search and understand, is very impressive,” says Oren Etzioni, former technical director of the Allen Institute for Artificial Intelligence, who was not involved in the work. “This is stuff I have not seen before.”

An AI that can operate across modalities would more closely resemble the way that human beings behave. “People are naturally multimodal,” Etzioni says, because we can effortlessly switch between speaking, writing, and drawing images or charts to convey ideas. 

Etzioni cautioned against taking too much meaning from the developments, however. “There’s a famous line,” he says. “Never trust an AI demo.” 

For one, it’s not clear how much the demonstration videos left out or cherry-picked from various tasks (Google indeed received criticism for its early Gemini launch for not disclosing that the video was sped up). It’s also possible the model would not be able to replicate some of the demonstrations if the input wording were slightly tweaked. AI models in general, says Etzioni, are brittle. 

Today’s release of Gemini 1.5 Pro is limited to developers and enterprise customers. Google did not specify when it will be available for wider release. 

Note: PLOS is delighted to once again partner with the Einstein Foundation Award for Promoting Quality in Research. The awards program honors researchers who reflect rigor, reliability, robustness, and transparency in their work. The Einstein Foundation received dozens of stellar submissions. We asked this year’s finalists to write about their research in the run up to the ceremony on March 14^th in Berlin. The is the first in our 5-part series.  

Author: Flavio Azevedo, Assistant Professor of Social Psychology, University of Groningen and his FORTT team (bios below).

Scientific research stands as a beacon of progress and innovation, yet its credibility has been threatened by repeated failure to replicate what had been widely believed. So, how can we ensure the reliability of scientific discoveries? This question lies at the heart of an ambitious project seeking to transform the landscape of scientific research, education, and policy.

The Challenge of Replication in Science

At the core of the scientific endeavor is the ability to accumulate knowledge via replication of studies—a fundamental yet often overlooked aspect. Replication involves repeating a study and testing a hypothesis under a variety of similar and different conditions to verify its results are reliable. It’s a litmus test for the robustness of scientific findings. Despite its importance, replication studies struggle with recognition, are less cited, not yet incentivized, and thus underutilized, leading to a skewed understanding of scientific literature. This has led to situations where studies are widely relied on before later replications show that the proposed effect was not real, or at least not generalizable

A Pioneering Solution: The FORRT Replication Database (FReD) Project

To address these issues, the Framework for Open and Reproducible Research Teaching (FORRT) has launched an ambitious project aiming to create a comprehensive, dynamic database cataloging scientific claims and subsequent replication attempts across various disciplines of social, cognitive, and behavioral sciences. This living, crowd-sourced collection not only includes replicated studies but also those that failed to replicate, providing a more complete picture of scientific inquiry.

Empowering Researchers and Educators

FORRT’s initiative extends beyond mere data collection. It involves developing user-friendly tools like Shiny apps, enabling researchers and educators to explore, visualize, and engage with replication data effectively. These tools will foster critical thinking about the planning, access, and integration of replication efforts in research and education.

Making Science Accessible and Credible

The project is set to revolutionize how we view, evaluate, and utilize scientific data. By making replication studies more accessible and understandable, FReD will empower educators and other stakeholders to incorporate these findings into their teaching and work, promoting discussions on the robustness of published research and reducing reliance on outdated evidence. Researchers will gain a valuable resource for generating new knowledge, and the public will access credible, reliable scientific evidence. The database is open to all and aligns with FORRT’s mission to democratize knowledge and higher education resources – in short, opening up science for everyone.

A Step Towards Transparent, Robust Science

The implications of FORRT’s work are profound. By bringing to light the successes and failures of past research, they lay the groundwork for a more transparent, robust, and reproducible scientific process. This is not just about data; it’s about building a culture of accountability, transparency, and integrity in scientific research.

In conclusion, the FReD project promises to enhance our pursuit of reliable knowledge. It’s a reminder that in science, as in life, understanding our past failures and successes paves the way for a brighter, more informed future.

Author biographies:

They’re versatile

One of my favorite things about battery technology is its adaptability. Researchers are finding and developing new chemistries all the time, and it’s fascinating to follow. 

Lithium-ion batteries tend to be the default for the industries I typically write about (think transportation and energy storage). That’s mostly because these batteries were developed for personal devices that became widespread beginning in the 1990s, so they’ve had a head start on scaling and the cost cuts that come along with it. 

Even in existing battery technologies, there’s lots of nuance and innovation. Lithium-ion batteries follow a similar blueprint, but there’s a whole world of flavors. Your phone and laptop probably house pouch cells with higher levels of cobalt, whereas your EV likely runs off cylindrical ones that are high in nickel. And a growing fraction of lithium-ion cells don’t include either of those metals—companies are looking at these options for stationary storage or lower- cost vehicles. 

But don’t stop there. Next-generation batteries could give us a different chemistry for every occasion. Need a robust, low-cost battery? Try sodium-ion. Even cheaper, for stationary storage? Zinc flow batteries or iron-air might be the chemistry for you. Something for a long-range, high performance EV? Check out solid state, or maybe something of the lithium-sulfur variety. 

I’m often asked which battery chemistry is going to “win.” Not all batteries are going to make it to widespread adoption, and not all battery companies are going to succeed. But I think the answer is that we’ll hopefully see not a single dominant type of battery, but an ever-growing menu of options. 

They’re at least a little bit magic

Last but not least, I think that one of the main reasons that I’m obsessed with batteries is that I find them a little bit mystifying. Tiny ions shuttling around in a metal container can store energy for us to use, whenever and wherever we want. 

I’ll never get sick of it, and I hope you won’t either. Here’s to spending more time with the ones we love in the year ahead. 

Related reading

Read more about lithium-sulfur batteries, which could unlock cheaper EVs with longer range, in my latest story. 

Jorge: Certainly. My role, I will call, has two major focuses in two areas. One of them is I lead the machine learning engineering operations of the company globally. And on the other hand, I provide all of the analytical platforms that the company is using also on a global basis. So in role number one in my machine learning engineering and operations, what my team does is we grab all of these models that our community of data scientists that are working globally are coming up with, and we grabbed them and we strengthened it. Our major mission here is the first thing we need to do is we need to make sure that we are applying engineering practices to make them production ready and they can scale, they can also run in a cost-effective manner, and from there we ensure that in my operations hat they are there when needed.

So a lot of these models, because they become part of our day-to-day operations, they’re going to come with certain specific service level commitments that we need to make, so my team makes sure that we are delivering on those with the right expectations. And on my other hand, which is the analytical platforms, is that we do a lot of descriptive, predictive, and prescriptive work in terms of analytics. The descriptive portion where you’re talking about just the regular dashboarding, summarization piece around our data and where the data lives, all of those analytical platforms that the company is using are also something that I take care of. And with that, you would think that I have a very broad base of customers in the company both in terms of geographies where they are from some of our businesses in Asia, all the way to North America, but also across the organization from marketing to HR and everything in between.

Going into your other question about how machine learning is helping our consumers in the grocery aisle, I’ll probably summarize that for a CPG it’s all about having the right product at the right price, at the right location for you. What that means is on the right product, their machine learning can help a lot of our marketing teams, for example, even when they are now with the latest generative AI capabilities are showing up like brainstorming and creating new content to R&D, what we’re trying to figure out what is the best formulas for our products, there’s definitely now ML is making inroads in that space, the right price, all about cost efficiencies throughout from our plans to our distribution centers, making sure that we are eliminating waste. Leveraging machine learning capabilities is something that we are doing across the board from our revenue management, which is the right price for people to buy our products.

And then last but not least is the right location. So we need to make sure that when our consumers are going into their stores or are buying our products online that the product is there for you and you’re going to find the product you like, the flavor you like immediately. And so there is a huge effort around predicting our demand, organizing our supply chain, our distribution, scheduling our plans to make sure that we are producing the right quantities and delivering them to the right places so our consumers can find our products.

Laurel: Well, that certainly makes sense since data does play such a crucial role in deploying advanced technologies, especially machine learning. So how does Kraft Heinz ensure the accessibility, quality and security of all of that data at the right place at the right time to drive effective machine learning operations or MLOps? Are there specific best practices that you’ve discovered?

Jorge: Well, the best practice that I can probably advise people on is definitely data is the fuel of machine learning. So without data, there is no modeling. And data, organizing your data, both the data that you have internally and externally takes time. Making sure that it’s not only accessible and you are organizing it in a way that you don’t have a gazillion technologies to deal with is important, but also I would say the curation of it. That is a long-term commitment. So I strongly advise anyone that is listening right now to understand that your data journey, as it is, is a journey, it doesn’t have an end destination, and also it’s going to take time.

And the more you are successful in terms of getting all the data that you need organized and making sure that is available, the more successful you’re going to be leveraging all of that with models in machine learning and great things that are there to actually then accomplish a specific business outcome. So a good metaphor that I like to say is there’s a lot of researchers, and MIT is known for its research, but the researchers cannot do anything without the librarians, with all the people that’s organizing the knowledge around so you can go and actually do what you need to do, which is in this case research. Never forget that data is the fuel, and data, it takes effort, it is a journey, it never ends, because that’s what is really what I would call what differentiates a lot of successful efforts compared to unsuccessful ones.

Laurel: Getting back to that right place at the right time mentality, within the last few years, the consumer packaged goods, or you mentioned earlier, the CPG sector, has seen such major shifts from changing customer demands to the proliferation of e-commerce channels. So how can AI and machine learning tools help influence business outcomes or improve operational efficiency?

Once in orbit, MethaneSAT’s software and spectrometers, which measure different wavelengths of light to detect methane, will pinpoint both concentrated locations for methane plumes as well as the broader areas where the gases diffuse and spread. It will also use Google’s image detection algorithms to create the first comprehensive, global map of the oil and gas industry’s infrastructure, like pump jacks and storage tanks, where leaks most commonly occur. 

“Once those maps are lined up, we expect people will be able to have a far better understanding of the types of machinery that contribute most to methane leaks,” says Yael Maguire, who leads geo-sustainability efforts at Google. 

This tool could solve a significant stumbling block for methane researchers, according to Rob Jackson, professor of Earth system science at Stanford. There are millions of oil and gas operations around the world, and information about where many of these facilities are located is tightly guarded, and where available, expensive to access. Some countries also block researchers from studying their infrastructure or using low-flying planes to measure emissions. With satellites, that may change.

“I think AI is the future of this field, where we should be creating databases of all these infrastructure types,” says Jackson, as measuring plumes from space sidesteps much of the oil and gas industry’s opaqueness on Earth. “One door that satellites are unlocking is the ability to peer everywhere. There will be nowhere to hide, eventually.” 

The MethaneSAT collaboration comes at a time when governments around the world are taking stronger stances on reducing methane leaks. Fueled by the momentum of COP28 in December, the Biden administration announced a new set of rules in December that will require more monitoring and repair of leaks. In January, the administration also proposed a fine against companies for excess methane, though that rule has not been finalized and is being fought by the industry. The European Union also agreed to stricter standards in November.

Lyten has made progress in stretching the lifetime of its batteries, recently seeing some samples reach as high as 300 cycles, Mickolajczak says. She attributes the success to Lyten’s 3D graphene material, which helps prevent unwanted side reactions and boost the cell’s energy density. The company is also looking to use 3D graphene, a more complicated structure than the two-dimensional variety, in other products like sensors and composites.  

Even with recent progress, Lyten is still far from producing batteries that can last long enough to power an EV. In the meantime, the company plans to bring its cells to market in places where lifetime isn’t quite so important. 

Since lithium-sulfur batteries can be extremely lightweight, the company is working with customers building devices like drones, for which replacing the batteries frequently would be worth the savings on weight, says Keith Norman, Lyten’s chief sustainability officer. 

The company opened a pilot manufacturing line in 2023 with a maximum capacity of 200,000 cells annually. It recently began producing a small number of cells, which are scheduled for delivery to paying customers later this year. 

The company hasn’t publicly shared which companies will receive the first batteries.  Moving forward, two of the company’s main focuses are improving lifetime and scaling production of both 3D graphene and battery cells, Norman says. 

The road to lithium-sulfur batteries that can power EVs is still a long one, but as Mikolajczak points out, today’s staple chemistry, lithium-ion, has improved leaps and bounds on cost, lifetime, and energy density in the years that companies have been working to tweak it. 

People have tried out a massive range of chemistry options in batteries, Mikolajczak says. “To make one of them reality requires that you put in the work.”

Lyten has made progress in stretching the lifetime of its batteries, recently seeing some samples reach as high as 300 cycles, Mickolajczak says. She attributes the success to Lyten’s 3D graphene material, which helps prevent unwanted side reactions and boost the cell’s energy density. The company is also looking to use 3D graphene, a more complicated structure than the two-dimensional variety, in other products like sensors and composites.  

Even with recent progress, Lyten is still far from producing batteries that can last long enough to power an EV. In the meantime, the company plans to bring its cells to market in places where lifetime isn’t quite so important. 

Since lithium-sulfur batteries can be extremely lightweight, the company is working with customers building devices like drones, for which replacing the batteries frequently would be worth the savings on weight, says Keith Norman, Lyten’s chief sustainability officer. 

The company opened a pilot manufacturing line in 2023 with a maximum capacity of 200,000 cells annually. It recently began producing a small number of cells, which are scheduled for delivery to paying customers later this year. 

The company hasn’t publicly shared which companies will receive the first batteries.  Moving forward, two of the company’s main focuses are improving lifetime and scaling production of both 3D graphene and battery cells, Norman says. 

The road to lithium-sulfur batteries that can power EVs is still a long one, but as Mikolajczak points out, today’s staple chemistry, lithium-ion, has improved leaps and bounds on cost, lifetime, and energy density in the years that companies have been working to tweak it. 

People have tried out a massive range of chemistry options in batteries, Mikolajczak says. “To make one of them reality requires that you put in the work.”