 |
|
|
|||||
Inside a lab at the Massachusetts Institute of Technology late last year, scientists gave an AI system a new task: designing entirely new molecules for potential antibiotics from scratch. Within a day or twofollowing a few months of trainingthe algorithms had generated more than 29 million new molecules, unlike any that existed before. Traditional drug discovery is a slow, painstaking process. But AI is beginning to transform it. At MIT, the research is aimed at the growing challenge of antibiotic-resistant infections, which kill more than a million people globally each year. Existing antibiotics haven’t kept up with the threat. The number of resistant bacterial pathogens has been growing, decade upon decade,” says James Collins, a professor of medical engineering at MIT. “And the number of new antibiotics being developed has been dropping, decade upon decade.” The research, recently published in the journal Cell, is part of his labs Antibiotics-AI Project and offers one example of AI’s potential in medicine. The team tried making a small number of the compounds, and then used one to clear a drug-resistant infection in a mouse. In another part of the study, the researchers used a different approach to generate additional molecules, leading to another successful test in miceand the possibility that novel, fully AI-designed drugs may eventually be available for the most dangerous infections. [Source Image: Walter_D/Adobe Stock] The current challenge The standard approach to creating new antibiotics involves screening an existing library of compounds, one by one, or sifting through samples of soil to find promising new candidates. Since the 1980s, the Food and Drug Administration has approved a few dozen new antibiotics, but most of them are minor variations on drugs that already exist. What’s happened in the last couple decades is, it’s largely been a discovery gap where folks are discovering antibiotics, but they’re more or less very similarand they are analogs to existing antibiotics, Collins says. The challenge is compounded by poor economics for drug companies. It costs effectively just as much to develop an antibiotic as it does a cancer drug or blood pressure drug, for example, he says. With an antibiotic, you might only take it once or only over a few days, whereas with a cancer drug or a blood pressure drug, you could take it for many months, years, or even for the rest of your life. With each use, an antibiotic also only makes a fraction of the profit.” All of this means that if youre infected with bacteria thats hard to treatlike methicillin-resistant Staphylococcus aureus (MRSA), which also resists many other drugsthere are fewer options available. In the U.S., MRSA kills an estimated 9,000 people each year. [Source Image: Walter_D/Adobe Stock] Evolving uses for AI The Collins Lab has been studying antibiotics for around 20 years. Initially, the team used machine learning to better understand how antibiotics work and to look for ways to make existing antibiotics more effective. Around six years ago, they started using artificial intelligence as a platform for antibiotic discovery. They used AI to screen existing libraries of compounds to look for new antibiotics, leading to the discovery of new molecules that worked against infections in new ways. A spin-off nonprofit, Phare Bio, is now working to move promising candidates toward the market. The biotech company hopes to launch a trial of halicin, a drug initially developed for diabetes treatment in 2009 that was discovered to have powerful antibiotic properties by Collins’s research team a decade later. The latest research goes a step furthernot just screening through existing compounds, but creating new ones. The scientists used two different approaches. First, they used a library of millions of chemical fragments known to have antimicrobial activity, and used the algorithms to turn those fragments into complete molecules. In the second approach, they used the AI to freely design new molecules, without starting from existing fragments. As the computer churned through new designs, the researchers were free to work on other tasks until the AI was done. After the molecules were generated, “we applied a series of down-selection filters to prioritize which ones to synthesize and test,” says Aarti Krishnan, a senior postdoctoral fellow in the lab. “Those steps took a few days and involved human feedback, where medicinal chemists manually inspected over 5,000 candidate molecules and selected them for synthesizability.” Actually making the molecules was challengingsome of the AI’s ideas were so wild that they would either be impossible or impractical to manufacture. (This will improve as the AI evolves.) But the team was able to make a small number. From the part of the study that worked from fragments of existing molecules, the scientists were able to make two candidates, one of which was very effective at killing drug-resistant gonorrhea bacteria. From the part of the study that let AI freely design new molecules, they synthesized and tested 22 samples, ultimately advancing one candidate in a successful test that treated drug-resistant MRSA in mice. Now, the lab’s nonprofit partner is continuing to work on both molecules so they can undergo more testing. [Source Image: Walter_D/Adobe Stock] A new use for generative AI While the use of AI in drug development isn’t new, this particular application of generative AI is. “To our knowledge, this is the first generative-AI approach that’s designed completely novel antibiotic candidates whose structures do not exist in any commercial vendor space,” Krishnan says. Drug development is still a slow process, and moving through human trials will continue to take time. But AI can clearly help in the early discovery phase, reducing cost and increasing the chances of success. “AI allowed us to explore much larger chemical spaces than are currently available from screening libraries. And in doing so, it opened up these new molecules for our consideration,” Collins says. The approach could also be useful for other types of medicine. “All of the AI methods that we use could be readily extended to other indications,” he says.
Category:
E-Commerce
AI image generators used to be terrible at handling text. Even once the models mastered hands with five fingers, the presence of mangled, nonsensical, vaguely Cyrillic-looking text was a dead giveaway that an image was generated by AI. Not anymore. Todays most advanced image generators have slowly improved their text generation. OpenAIs image generator within ChatGPT handles basic text tasks fairly well. And design-centered models like Ideogram are great for simple, practical text tasks like creating video thumbnails. This week, though, Google has released Nano Banana Pro, an updated version of its wildly popular AI image editing tool. Nano Banana Pro, like its predecessor, is middling when it comes to generating realistic AI photographs. But its absolutely amazing at creating beautiful, informative, accurate infographics. In fact, the model is so good that it can turn literally anything into a professional quality infographic in a matter of seconds. Dont Read To Me Personally, I absorb very little of what I hear. Im a visual learner, so if I listen to a presentation without taking physical notes or seeing some kind of visual aid, most of the information dissipates into the ether before my brain has any shot at absorbing it. Add background noise or a presenter who mumbles, and Im totally screwed. I love infographics because they take complex information and lay it out in a way my brain can grok. I can glance at a graphic and absorb more information than Id get in a two-hour lecture. In my testing, I was therefore thrilled to see Nano Banana Pros remarkable ability to take pretty much anything I threw at it, and turn the data into a bespoke visual aid. First, I started with some practical use cases. I fed Nano Banana Pro data on the recent performance of a YouTube channel Ive been developing called California Dad Reviews. Based on my analytics data, Nano Banana spun up an infographic showing whats doing well on my channel, several standout videos from the last month, and its recommendations for what to shoot next. In this case, I gave the model unique data from my channel. But because Nano Banana Pro is integrated within Googles Gemini 3 chatbot, it can also perform background research on its own. In another test, I asked the model to research the best times to cross the San Francisco Bay Bridge, and then present its results visually. Its impressively comprehensive and useful. Graphics Get Personal These kinds of informative graphics are helpful. Im sure that bloggers will be tripping over themselves to integrate Nano Banana Pros infographics into their SEO optimized posts. But in my testing, I found that the best uses for the tools new visual capabilities are more personal. I recently took my senior dog, Lance, to the vet for his annual checkup. He got some routine bloodwork, and his vets report was laden with medical terms and specifics. I fed the whole thing into Nano Banana Pro, and got a clean infographic summarizing the findings. Spoiler alert: hes doing great! The model is especially powerful because its able to process nearly any kind of input data. In planning a day trip for a large group in downtown San Francisco, I copied and pasted a long WhatsApp chat with lots of logistics into the model. It spat out a pretty graphic summarizing the days plans, complete with a bespoke map of the city. I shared it to the group, and people loved it. Show me the Visuals Again, as a visual learner, the ability to conjure up an infographic in a few seconds (and for free) from essentially any input data is incredibly valuable. Its also easy to do. You open the Gemini app, paste in the data you want to process, select the Create Image option, and let Nano Banana Pro plug away. Beyond dg health visualizations, though, Googles new model says a lot about where visual AI is going. In the early days of image generators, creating fun, bizarre images (Ballerina Cappuccina, anyone?) was enough to hold users attention. You can only generate so many AI cat photos or Hunky Jesus memes before the tech gets old, though. In response, AI companies are increasingly specializingcreating image generators that solve bounded, real-world tasks. Again, tools like Ideogram are tailored to designers. Adobe has a whole suit of generative tools built into its iconic Photoshop and Premiere products for photographers and videographers. And ChatGPTs models are perfect for things like making event posters. For Google, though, the endgame has always been about processing tons of information and summarizing it for users. We saw that in the companys classical 10 blue links, and more recently in its wildly popular AI Overviews. Now, Google appears to be using its AI image generation prowess to summarize and present information visually. That evolution means were almost certain to see Nano Banana generated infographic images appear within other Google toolsfirst AI ones like NotebookLLM, and later within live search results presented to everyday users. With my brains preference for visuals, Im thrilled to see this new direction play out. To get the ball rolling, I fed this entire article into Nano Banana Pro. True to form, the infographic is beautiful:
Category:
E-Commerce
Below, Tim Elmore shares five key insights from his new book, The Future Begins with Z: Nine Strategies to Lead Generation Z as They Disrupt the Workplace. Elmore is the founder and CEO of Growing Leaders, a nonprofit dedicated to developing emerging leaders. As a speaker and coach, he has helped organizations from universities to Fortune 500 companies connect more effectively across generations. Whats the big idea? Many leaders are scratching their heads over Gen Z. The old playbook doesnt work anymorebut figuring out how to engage and collaborate with this generation is what turns good leaders into great ones. Listen to the audio version of this Book Biteread by Elmore himselfbelow, or in the Next Big Idea App. 1. Turn frustration with Gen Z into hope According to a survey by Resume Builder, three in four managers agree that todays young employees, from Generation Z, are the toughest generation to lead. I believe Gen Z is frustrating to us because they seem so very different, and often unexplainable. I think its because of a phenomenon I call the Peter Pan Paradox. Peter Pan could mysteriously fly into London and sprinkle Pixie dust everywhere to make magical things happen. On the other side of the coin, Peter Pan wanted to live in Neverland, where boys never have to grow up. Over the last decade, Ive noticed something magical and tragic happening in culture. The Age of Authority is decreasing. Gen Zers often come in with an authority that doesnt require a title. They are intuitive about using AI. They see where culture is going. They know how to monetize social media, and they have visibility on future customers. At the same time, the Age of Maturity is increasing. They often come in behaving unprofessionally. The pandemic delayed their growth. Their social and emotional skills are delayed in development. Employers are saying 26 is the new 18. Almost one in three employers fire them in the first month. So, how should leaders respond? We must listen more than we used to listen to young staff. We must coach more than we used to coach young staff. Maya was hired right out of collegeand immediately her team could tell shed never had a full-time job before. She wore flip flops to work and often arrived at the office 10 minutes late. Maya spent most team meetings quietly, glancing at her phone from time to time. Her manager spent extra time coaching her on the fundamentals, hoping shed catch up. Then, one day, the payoff happened. Maya burst into her managers office with a big smile on her face. She told him shed been thinking about a problem their team had been discussing for months. Maya had an idea on how they could use AI to solve that problem. It ended up being an incredible solution. No one had known what was going on inside her head. You might say, Maya became her own version of Peter Pan. 2. Gen Z ROI is well worth it In 2018, Colin Webb graduated from the Massachusetts Institute of Technology. While at MIT, he served four terms as class council president and now serves his alma mater as a member of the MIT corporation, the schools board of trustees. Upon graduation, he was offered a job by General Motors. Colin was asked to serve as a design, release, and development engineer for their Cruise autonomous vehicle program. He helped make smart cars. He quickly realized he was part of an old industry with a traditional style of getting things done. Obviously, he and his young teammates brought some new ideas, but when he bounced them off his supervisor, he was told to keep his head down and his nose to the grindstone. Soon after, however, Collin did something audacious. He emailed Mary Barra, the CEO of General Motors, and shared his ideas for improving the company. Mary replied and agreed that his ideas were good and that shed take them to her executive team. They, too, agreed his ideas were good. But when those ideas made their way down the organizational chart, they died on the vine. Again, Collin was told to keep his head down and nose to the grindstone. In fact, he was told it would take about eight years before hed get the chance to lead anything. They might as well have told Collin to leave. Within a year, Collin had left the company, even though he is not a quitter. Since then, he has started three companies and sold one of them. Hes doing very well, working hard and living his dreamworking with AI. Sadly, similar stories happen all the time. Too often, supervisors miss what Gen Zers bring to the team for three reasons: When were comfortable, we default to: Thats not the way we do it here. When were scared, we become more concerned with procedures than with progress. When were experienced, we assume theyre young and dont know much. In 2017, a woman walked into a New Hampshire thrift store looking for a picture frame she could restore. She found one for four dollars and thought, Even if I cant use it, the price is inexpensive. She bought it. When she got home, her family examined it and said, The frame is nice enough, but the painting inside looks like its from a famous artist. You should have it evaluated. When the woman did, she discovered it was an N.C. Wyeth painting that later sold for $191,000. Not a bad trade-off. In many ways, this is an analogy for Gen Z. We usually dont spend a lot on them right away since all we can see is their frame. Later, we realize its whats inside of them thats so valuable. Lets stop treating them like expendable commodities and begin treating them like currency worthy of investment. Lets find a way to connect with and get curious about them. This may be key to thriving in the future. 3. Relationship and trust mean everything to Gen Z Many Gen Z team members enter their careers directly after college, without having had a full-time job before. Parents encouraged them to focus on academics, but we all know that the classroom rarely resembles the office. Four in five Gen Z job candidates go so far as to bring their parents with them to the interview. At the same time, they also bring with them a new perspective on authority. Much like Generation X some 35 years ago, they are suspicious of corporate leaders, political leaders, educational leaders, and even religious leaders due to stories of corruption. This has fostered a significantly different view of authority than that held by seasoned business veterans. OLDER GENERATIONS Position gives you right to influence. Older folks have wisdom. Systems offer order to chaos. We must listen to the man at the top. The top dog wins the debate. The leader is a gatekeeper. YOUNGER GENERATIONS Connection gives you right to influence. Older folks may be irrelevant. Systems must be disrupted or grow. Top people should be listening to us. The best idea wins the debate. The leader is a guide. Relationships and trust mean everything to most members of Generation Z. While research confirms they dont trust most traditional institutions, they do trust peopleeven people from older generationswhom they get to know. So, here are the shifts we should make: Dont think control, think connect. Dont think inform, think interpret. Dont think what, think why. Dont think inputs, think outcomes. Dont think tell, think ask. Dont think manage, think mentor. We dont know much about the future, but we do know that Generation Z will be there. Lets build bridges of relationship that can bear the weight of honest disclosure. Everyone will win. 4. How to hire Gen Z Thousands of companies are struggling with the diminishing number of job candidates. I describe what lies ahead in a metaphor I call A New Kind of Storm: We are facing a blizzard. Young people often wont take traditional entry-level jobs as they did in the past. This blizzard is part of a new winter season. A smaller population, plus issues with addiction and incarceration, reduces the employment pool. This may be part of a larger new ice age. Within a decade, there will be more 65-year-olds than 16-year-olds in the U.S. So, what can we do? Make your organizations offering more attractive. Make their job the best first job to launch their career. Approach your role as a mentor, not a manager. Gen Z seems to know its a buyers market when it comes to getting a job. Sally hosted an interview with a Gen Zer (well call him Owen). In the interview, she asked Owen to describe a time when he faced a challenge and was proud of how he made it through. She was pleased with his response and, three days later, asked HR to extend an offer to Owen. His response? He said, Are you kidding me? Theres no way I will work for that woman. She triggered my PTSD when she asked me to talk about a challenge of mine. It didnt feel psychologically safe. We need a new approach to job interviews, especially with Gen Z candidates. For many, its a new experience, and employers can use this acronym to get acquainted: PPreferences: What do you wish to be true about the job?EExpectations: What are your assumptions about the work?RRequirements: Any demands or deal breakers?KKeys to their heart: Get to know them beyond their role as a worker.SSalary: Clarify, and possibly negotiate. This is your chance to build a bridge. Dont miss the opportunity to do so. 5. Not all Gen Z myths are true Many bosses (or people over 45) see Gen Z behave in certain ways, dont understand it, and then stereotype them. During my research for this book, I shared some of these stereotypes with Gen Z focus groups. I will never forget meeting with one group and communicating that many leaders believe that Gen Zers dont really want to work. After all, they leave right at 5 p.m., and dont stay a minute longer. They often leave a task halfway done and head out. It feels like they dont like to be there. One 21-year-old, full-time team member replied, Dr. Tim, do you know why I leave right at 5? I rush out because I need to get to another job. I dont make enough to pay the bills at this first job. Then, I go take care of my mom, who has stage 4 cancer. Suddenly, I realized this Gen Zers work ethic is just fine. There is, however, one stereotype thats not a myth. Millions of Gen Zers can be fragile, especially when receiving feedback. It seems that hard is the new harsh. Some have never had to hear firm feedback face-to-face with a boss. One of the most challenging situations leaders must face is hosting difficult conversations. Many supervisors avoid them. I created another acronym for sitting down with and correcting a teammate: AAsk: Asking questions makes them feel valued.LListen: Listening makes them feel heard. Gen Zers want to have a voice.EEmpathize: Empathizing makes them feel understood.GGuide: This is when you can provide the necessary feedback. The correction was communicated using a bridge, not a badge. Enjoy our full library of Book Bitesread by the authors!in the Next Big Idea App. This article originally appeared in Next Big Idea Club magazine and is reprinted with permission.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||