Summary: Moderna goes all-in on AI use | Neopets revived after 20 years: pets still live | Improving the usability of empty-state areas in a UI | AI scores well on medical exams compared with human doctors | Behind the scenes in Jakob’s office when recording or broadcasting webcasts | Free live keynote by Jakob Nielsen this Wednesday | Mentorship from the perspective of the mentor and the mentee
UX Roundup for May 13, 2024. (Midjourney.)
Moderna All-In on AI
Moderna is one of the world’s most innovative pharmaceutical companies. Their most famous achievement was needing less than one month between receiving the DNA code for the Covid-19 virus (January 11, 2020) and manufacturing the first batch of the vaccine for clinical testing (February 7, 2020). It then took the government 10 more months to allow the general public to get the vaccine (authorized December 18, 2020), but the invention was fast.
Moderna continues to live up to its name, by being a leader in adopting AI for drug discovery. A nice 3-minute video gives an overview of AI at Moderna with comments from the CEO and CIO (free on YouTube), with some more details in a Wall Street Journal article (subscription required). Brad Miller, CIO, emphasizes the importance of integrating AI with the workflow and transforming the workforce around the new capabilities. In a few months, Moderna designed 750 custom GPTs for its specific work. (“GPTs” is OpenAI’s stupid name for a customized version of its foundation model.) One such GPT, called “Dose ID,” uses years of previous research and medical data to predict the optimal dose of a drug for clinical trials, which is apparently difficult to do and has caused trouble in previous clinical trials.
Even though drug discovery is the main use, Moderna’s Chief Legal Officer Shannon Klinger boasts that the legal department was the first to achieve 100% adoption of AI. (She doesn’t define “100% adoption,” but I assume it means that all her staff uses AI in some form or other, such as summarizing a contract they’re reviewing.)
The CEO states that he expects to launch 15 new drugs in the next 5 years, which might have required as many as 100,000 new staff to accomplish in an old-school pharma company without AI. But now, because of AI's productivity improvements, he will only need to hire a few thousand extra staff. While it’s nice for shareholders to save that money, it’s even more important for humanity to expedite drug innovation, which will absolutely save lives.
AI expedites drug discovery at Moderna. (Midjourney)
Neopets Revived After 20 Years: Pets Still Live
Neopets was a popular website during the dot-com years, especially among children. The service was bought by a bigger company and left to falter without upgrades, as the world passed it by, with more exciting simulation games from FarmVille to Minecraft.
However, for the entire 25 years, Neopets stayed up and running. Neopets is now owned by an independent company, dedicated to its success, and development has restarted: the Neopets user base has tripled in the last 6 months.
Of particular interest is that many Millennial users who were fans as children 20 years ago have returned to Neopets as adult business professionals and reconnected with the virtual “pets” they left behind.
As I pointed out in my review of the time scales of UX, some UX decisions last for decades, and one of the longest-lasting UX elements is user data. In this case, the new World of Neopia company inherits the legacy of the long-gone team who designed the original Netpets to keep the pets alive even during decades of absent users.
Grownup users return to Neopets and discover that their childhood virtual “pets” are still alive on the site after 20 years of absence, showing the value of designing for long-lasting user data. (Ideogram)
Improving the Usability of Empty-State Areas in a UI
Megan Legawiec has a nice overview of how to handle the “empty state” of a user interface: when there’s no content or items to be displayed in a list or other area that would usually be populated.
She has 6 tips, but I’ll just give you the 3 most important ones:
Active voice: say what can the user do here, since there are no examples to show the way.
Help users get back on track and avoid blaming them for the lack of items.
Tell the user when things are going well (e.g., all errors have been cleared, which means the list of errors is empty).
Emptiness can be intimidating, maybe even depressing. In a UI, showing things usually helps users understand what they can do, but when there’s nothing to display, the UX copy must come to the rescue. (Midjourney)
AI Doctor Better than No Doctor and Better than Many Human Doctors
New research from the Tel Aviv University School of Medicine indicates that AI already performs close to the level of human doctors — at least when it comes to exam scores.
Percentile scores for AI taking official board residency examinations, relative to the scores received by human doctors taking those exams. The percentile indicates how many human doctors were outscored by AI. So, for example, on the psychiatry exam, GPT 3.5 scored better than 13% of the human doctors (and thus worse than 87% of human doctors), whereas GPT 4 scored better than 75% of the human doctors. Only 25% of human doctors taking the psychiatry exam were better than AI.
| GPT 3.5 | GPT 4 |
Pediatrics | 2% | 17% |
Obstetrics/Gynecology | 0% | 23% |
General Surgery | 0% | 44% |
Internal Medicine | 5% | 57% |
Psychiatry | 13% | 75% |
Average | 4% | 43% |
These are the results of AI taking traditional, written exams. Results are worse when testing AI’s ability to diagnose patients based on scans, for example, looking at endoscopy images, chest x-ray, and skin lesions. Currently, language models are better than image recognition models. I’m sure both classes of AI will improve.
Three important conclusions from this data:
Never use GPT 3.5 (the free version of ChatGPT) for anything important. GPT 4 is immensely better. Once GPT 5 is released, my advice will likely be updated to also avoid GPT 4, though it’s currently not known how much better 5 will be.
AI is still not as good as the average human doctor, across disciplines. In psychiatry, though, AI is already better than most humans, possibly because this discipline is so strongly based on language.
AI is likely already good enough that it’s better than no doctor. Poor or no medical care is the standard in much of the world, and in those regions, providing access to health care through AI will likely save millions of lives per year.
Medical care from an AI is better than what’s available in many locations with no or substandard physicians. And it may already be better than average human doctors in some specialties. (Ideogram)
Behind the Scenes in Jakob’s Office When Recording or Broadcasting Webcasts
This is a photo of my home office setup when recording webcasts and podcasts. I hang a big sheet of brightly colored paper behind my chair, which creates a very professional studio image when broadcast through my webcam. I hang different colors for different shows, for an easy and cheap way of varying the look of my events.
Here, the setup is seen from the back. Behind the paper is my computer with a webcam and a high-end Blue Yeti microphone. Much research shows that audio quality is more important than video quality when people judge a speaker.
My colored paper is a roll from Savage. It’s cheap and great. The stand, on the other hand, is an expensive model from Manfrotto, which is also the brand of my two light stands (front light and side light). Don’t buy cheap, flimsy stands: there is too much risk of things falling over in mid-broadcast. Finally, to the right, you will see the sheet music stand I repurposed to hold my notes when giving keynote speeches. (I don’t use it when I’m on a webcast.) If you look very closely, just to the left of the music stand is one last essential component of good broadcasting from a home office: wired ethernet cable. Wi-Fi is good for many things, but it drops video packets.
Real photo.
Live Keynote by Jakob Nielsen This Wednesday: 10 Foundational UX Insights
I will present a live keynote on ADPList this Wednesday (May 15, 2024) in support of their mentorship mission on the topic of “10 Foundational UX Insights.” I thank my good friend (and stellar UX researcher) Alita Joyce from Google for being the session chair and managing the Q&A.
The event is free, but advance registration is required.
My speech is live on the Internet at these times:
San Francisco: 10 AM USA Pacific Time
New York: 1 PM USA Eastern Time
São Paulo: 2 PM Brazil Time
London: 6 PM British Summer Time
Paris/Berlin: 7 PM Central European Time
Dubai: 9 PM Gulf Standard Time
New Delhi: 10:30 PM India Standard Time
Singapore/Beijing: 1 AM the next day (May 16) Singapore Time/China Time
Here’s the summary of the talk:
UX has come a long way since its early beginnings at Bell Labs in the 1940s. As we enter Year 2 of the AI revolution, it’s clear that the core principles of UX design are not being replaced but are evolving with AI integration. This journey through the 10 foundational insights of UX reminds us that while “there's nothing new under the sun,” the way we apply these insights is constantly changing. By embracing AI as a tool for enhancing these time-tested principles, we can create more personal user experiences than ever before. Let’s honor the past as we design for the future, ensuring that the essence of UX — empowering humans while making technology subservient — remains at the heart of everything we do.
Jakob Nielsen presents a live keynote on ADPList on this Wednesday (Thursday morning in East Asia/Australia). (Midjourney)
Mentorship: Mentor’s and Mentee’s Perspective
Christine Sheller, Senior Vice President at O3, wrote a nice article on the power of mentorship in UX. She covers both sides of the relationship. Her key recommendations (other than that you should have a mentor or mentee, respectively) are:
Define goals for the relationship
Prepare for meetings with specific topics or questions
Be open to feedback
Follow up on action items discussed
I believe that mentors are now the best way to learn UX, other than the specific low-level skills you can learn by just-in-time advice from AI. (Thus, I also advise against wasting a senior mentor’s time to ask about something you can learn from Perplexity.) Will AI be able to provide more strategic advice and help navigate the human side of advancing in the UX profession? Maybe one day. I only say “never AI” for a few things. But for now, mentors are the game, particularly because ADPList has made it easy to find mentors with all kinds of expertise from all corners of the world.
In trouble at work? Getting a more experienced mentor’s perspective can help. (Midjourney)