Emotions and Tech/ CRISPR Revolution/ AI Determined Beauty/ Harari’s Lessons from a Year of Covid/ Driverless Future

This week I met my grandfather. This wouldn’t be worth noticing, if not that he passed away in 1952, when my father was sixteen. I had heard a lot of stories about him, and how much he and I had in common, both physically and character-wise. But the only thing I had about him was an old picture, taken in his early forties.

After writing about “deep nostalgia” last week, I decided to go for it myself. I asked my father for the “one” picture of my grandfather, based on which my whole imagination about him built over the years. I uploaded it, and there he was!

It has been a very fascinating experience. Rationally, I was very well aware that this was simply a rendering of an old picture, made using a generative algorithm, i.e. pure “cold” technology. Emotionally though, it was me meeting for the first time my grandfather, “rendered” through an algorithm and therefore animated by it. It was a very emotional moment for me. After decades thinking of him, and how he could have looked like, beyond that one picture, I finally could “meet” him. Yes, it was only a 17 seconds video, but for me it was much more that, it was the becoming real of a long series of thoughts while imagining how he could really look like.

I also couldn’t stop thinking about the irony that behind turning a photo into a live memory that felt “real” to me was what we call “artificial” intelligence. I also shared the video and the experience with my wife, and her first reaction was that it was “creepy”. At the same time, I shared the video with my father, who last saw an animated picture of his father almost 70 years ago, and also for him it was a highly emotional moment.

I have been thinking about this since last Sunday, when I had the “encounter”. I have been really touched by the emotional experience and realized how much we are underestimating this dimension in all our discussions around AI (and technology in general). I knew that what I was seeing was not real, but this was now what and how I was feeling. That WAS my grandfather, that is what I was feeling, and somehow, I could build an emotional relationship with him through that video, regardless of everything my rational brain was telling me. 

As we move into the future, let’s not forget the emotional side, for all technology power. How we are going to feel is going to be at least as important as we think. It is probably a platitude, but one worth being remembered of.

I Asked an AI to Tell Me How Beautiful I Am

A quick Google search will yield a bevy of startups around the world - from China to Japan to Australia - offering facial recognition and analysis APIs leveraging AI for use cases like beauty scoring, makeup recommendations, dating apps, and more. The world’s largest open facial recognition platform, Face++, has a beauty scoring AI. Large US companies have also invested in beauty AI: "Ulta Beauty, valued at $18 billion, which developed a skin analysis tool. Nvidia and Microsoft backed a robot beauty pageant in 2016, which challenged entrants to develop the best AI to determine attractiveness."

When you're putting percentage points on something as subjective as attractiveness, you have to ask what's under the hood. Cosmetic surgery consultation startup Qoves claims it's more forthcoming about its facial analysis system: "Founder Shafee Hassan claims that beauty scoring might be even more widespread. He says that social media apps and platforms often use systems that scan people’s faces, score them for attractiveness, and give more attention to those who rank higher. 'What we’re doing is doing something similar to Snapchat, Instagram, and TikTok,' he says. 'but we’re making it more transparent.'"

But experts argue that companies should be even more transparent. Serge Belongie, a computer vision professor at Cornell University, says companies "should own it and say yes, we are using facial beauty prediction and here’s the model. And here’s a representative gallery of faces that we think, based on your browsing behavior, you find attractive. And I think that the user should be aware of that and be able to interact with it."

News items:

How Europe's €100 Billion Science Fund Will Shape 7 Years of Research

Horizon Europe, the world’s largest multinational research and innovation programme, has issued its first call for grant applications.

"The Code Breaker": Jennifer Doudna and How CRISPR May Revolutionize Mankind

Some 7,000 human diseases are caused by gene mutations. Emmanuelle Charpentier and Jennifer Doudna's Nobel Prize-winning CRISPR technology, introduced only eight years ago, may hold the key to eradicating them. CRISPR is an extremely versatile tool. Scientists have already used CRISPR to breed more nutritious tomatoes, create a wheat that doesn't contain gluten, and conduct clinical trials to treat some cancers. In 2019, Vicotria Gray became the first person to be treated for sickle cell disease using CRISPR: "In the year since receiving the experimental treatment, she's had no severe pain or hospitalization."

The technology, of course, has a darker side. As many sci-fi movies and 2019's CRISPR baby scandal have shown, powerful but simple genetic engineering technology can be used in unethical, even dangerous ways. But Doudna assures us that "we don't really know which genes need to be edited.... I suspect that we're talking about dozens, if not more.... that would be technically very challenging. So, I don't think we're on the verge of a world of CRISPR babies myself."

Walter Isaacson, whose recent book The Code Breaker chronicles Charpentier and Doudna's CRISPR efforts, notes that "most people who have studied this say you got to draw a line between what's medically necessary – in other words, trying to make sure people don't get sickle cell anemia or Huntington's – but it's a blurry line. I mean, if you're trying to improve somebody's memory to make sure they don't have Alzheimer's, you're also improving their memory."

News items:

How Will the Proteins of the Future Be Designed?

Biomatter Designs, a Lithuanian startup is working to create a next-gen protein design platform through synthetic biology and artificial intelligence.

Could Fruit Flies Help Match Patients With Cancer Treatments?

A British company is seeding genetically modified flies with human tumors, giving patients their own personal drug trials.

Building AI for the Global South

In the past few years, there has been a surge of research papers exploring efforts to create ethical AI guidelines not centered around Western ideas of fairness. One recent paper, titled Re-imagining Algorithmic Fairness in India and Beyond, was submitted to this week's Conference on Fairness, Accountability, and Transparency, and urges researchers not to "copy-paste the Western normative fairness everywhere" when creating AI for the Global South. ("A 2019 paper about designing AI for the Global South describes the term 'Global South' as similar to the term 'third world,' with a shared history of colonialism and development goals.")

The paper was co-authored by Google research scientists Nithya Sambasivan, Ben Hutchinson, and Vinodkumar Prabhakaran. The latter two of which were co-authors on the Dangers of Stochastic Parrots: Can Language Models Be Too Big report - the paper at the center of Timnit Gebru's firing. In Re-imagining Algorithmic Fairness, the authors conducted 36 interviews with professionals working with marginalized communities in India. The paper cites the AI Observatory, a project that's taking steps in the right direction by documenting harm from automation in India, and also "calls for reporters to go beyond business reporting and ask tech companies tough questions."

News:

In Battle With U.S., China to Focus on 7 'Frontier' Technologies From Chips to Brain-Computer Fusion

China is looking to boost research into what it calls 'frontier technology' including quantum computing and semiconductors, as it competes with the U.S. for supremacy in the latest innovations.

Facebook’s New AI Teaches Itself to See With Less Human Help

Facebook has shown how some AI algorithms can learn to do useful work with far less human help.

Engineering a Driverless Future

"In the next 10 years we're going to see the impact of safety [that autonomous cars will bring to] society. We forget the status quo. The 38,000 in US, 1.2M ppl globally that die in accidents. We can start to put a dent in that as these vehicles are deployed." - Chris Urmson

Azeem Azhar recently had a great conversation with co-founder and CEO at self-driving startup Aurora, Chris Urmson, a humble engineer who has been involved with autonomous vehicles since DARPA issued their Grand Challenge in 2004. Besides having some fun anecdotes about driving through fence posts in the Mojave, Urmson explains his "no jerks" policy at Aurora (he believes the idea that you should hire 10x engineers even if they aren't team players is ridiculous), why the company is fine aiming for Level 4 autonomy (Level 5 would be able to drive in all conditions), and why, if you want to get to the moon, you build a rocket, not a ladder...

"If you want to get to the moon, it's really appealing to build a ladder. But the problem is, you're not going to get to the moon building a ladder, even if it looks like you're getting close every day. In contrast, our approach has been to build a rocket. Let's build those foundational technologies that sit on the pad for a while looking like you're not going anywhere. But as it combines, suddenly [a day passes] and you're there."

News items:

Don't Swat This Bug. It Might Be a Robot on a Rescue Mission

Kevin Chen, an assistant professor at the Massachusetts Institute of Technology, envisions a time when his insect-sized drone could be used as a search and rescue robot.

How to Build an Artificial Heart

Daniel Timms started working on his artificial heart in 2001, when he was twenty-two years old.

Yuval Noah Harari: Lessons From a Year of Covid

"Epidemics are no longer uncontrollable forces of nature. Science has turned them into a manageable challenge. Why, then, has there been so much death and suffering? Because of bad political decisions." - Homo Sapiens author Yuval Noah Harari

At no point in history were humans as prepared for a pandemic as 2020. If you instituted nationwide lockdown measures during the 1918 influenza pandemic, it would result in "economic ruin, social breakdown and mass starvation," since a good portion of the labor force was working in agriculture, movements of pre-symptomatic or asymptomatic carriers couldn't be traced, and a myriad of other factors. Ten days into 2020, scientists had already sequenced the virus's genome, and within a year vaccines were in mass production. Trade, once deadly due to the amount of humans involved in long-distance travel, is relatively safe today: "A largely automated present-day container ship can carry more tons than the merchant fleet of an entire early modern kingdom. In 1582, the English merchant fleet had a total carrying capacity of 68,000 tons and required about 16,000 sailors." In 2020, global maritime trade declined by only 4%.

For all the amazing progress we've made in circumventing a deadly virus, "science cannot replace politics." There is no scientific way of measuring interests and values. When imposing a lockdown, writes Harari, you don't just ask how many people will avoid infection - you ask "how many people will experience depression if we do impose a lockdown? How many people will suffer from bad nutrition? How many will miss school or lose their job? How many will be battered or murdered by their spouses?" Asking what really counts is a political task. Unfortunately, political responsibility wasn't a top priority in 2020: "The early months of 2020 were like watching an accident in slow motion.... no global leadership emerged to stop the catastrophe from engulfing the world. The tools have been there, but all too often the political wisdom has been missing."

News items:

Imaginary Numbers May Be Essential for Describing Reality

A new thought experiment indicates that quantum mechanics doesn’t work without strange numbers that turn negative when squared.