Afterglow of The Big Bang.
A different kind of Weekend Edition.
1. First things first:
Biotech startup Loyal has raised $45 million in equity to help bring its first longevity drug for dogs to market.
Bain Capital Ventures led the Series B financing, with participation by new backer Valor Equity Partners, as well as returning investors Khosla Ventures, First Round Capital and others. Bain Partner Kevin Zhang has joined Loyal’s board. Bridge Bank, part of Western Alliance Bank, has also provided a $12.5 million line of credit to the company.
Loyal’s valuation increased, said founder and chief executive Celine Halioua, but she declined to comment on specific figures. Research firm PitchBook Data put the San Francisco startup’s valuation in its prior round in 2022 at $220 million. (Source: wsj.com)
2. Cosmologists are preparing to cast their sharpest-ever eyes on the early Universe. From an altitude of 5,300 meters on Cerro Toco, in northern Chile’s Atacama Desert, the Simons Observatory will map the cosmic microwave background (CMB) — sometimes called the afterglow of the Big Bang — with a sensitivity up to ten times greater than that of the previous gold standard, Europe’s Planck space probe. “It will be the best view of the CMB that we’ve ever had,” says Jo Dunkley, a cosmologist at Princeton University in New Jersey and one of the leading researchers in the observatory’s team. Construction of the US$109.5-million observatory is due to be completed in a matter of weeks. One of the project’s goals is to find fingerprints left in the CMB by gravitational waves that originated from the Big Bang itself. These would provide the first incontrovertible evidence for cosmic inflation, a brief moment in which expansion is thought to have proceeded at an exponential rate. (Source: nature.com)
3. An instrument on a NASA spacecraft due to blast off to Europa later this year may be able to directly detect cellular material ejected from the icy moon of Jupiter, raising the prospects for finding life. Europa has garnered scientific interest because researchers believe it contains a vast, saltwater ocean under its thick icy shell. It is also surrounded by an orbiting blanket of ice grains and dust, believed to be remnants of material thrown up following bombardments by meteorites. NASA’s Europa Clipper spacecraft, due to launch in October and arrive at its destination in 2030, will fly near the moon, but won’t land on it. It will carry 10 experiments with the aim of studying Europa’s internal structure, including the chemistry of its ocean and its potential habitability for life beyond Earth. (Source: newscientist.com)
4. The abstract from Sakana AI’s research report, Evolutionary Optimization of Model Merging Recipes:
We present a novel application of evolutionary algorithms to automate the creation of powerful foundation models. While model merging has emerged as a promising approach for LLM development due to its cost-effectiveness, it currently relies on human intuition and domain knowledge, limiting its potential. Here, we propose an evolutionary approach that overcomes this limitation by automatically discovering effective combinations of diverse open-source models, harnessing their collective intelligence without requiring extensive additional training data or compute. Our approach operates in both parameter space and data flow space, allowing for optimization beyond just the weights of the individual models. This approach even facilitates cross-domain merging, generating models like a Japanese LLM with Math reasoning capabilities. Surprisingly, our Japanese Math LLM achieved state-of-the-art performance on a variety of established Japanese LLM benchmarks, even surpassing models with significantly more parameters, despite not being explicitly trained for such tasks. Furthermore, a culturally-aware Japanese VLM generated through our approach demonstrates its effectiveness in describing Japanese culture-specific content, outperforming previous Japanese VLMs. This work not only contributes new state-of-the-art models back to the open-source community, but also introduces a new paradigm for automated model composition, paving the way for exploring alternative, efficient approaches to foundation model development. (Sources: sakana.ai, arxiv.org/abs, italics mine)
5. How Big Tech is winning the AI talent war:
Microsoft’s latest hiring spree, following its $13 billion investment in OpenAI and its more recent partnership with France’s Mistral, highlights the company’s intent to ally with ambitious AI start-ups and command the market.
The company’s hyperactivity has helped Microsoft re-emerge as the world’s most valuable public company with a market value of $3.1 trillion, more than all the companies listed on London’s FTSE 100 index put together. Once again, Microsoft has left its arch-rival Google spitting dust.
Even relatively junior engineers at the leading research companies that are developing AI foundation models, including OpenAI, DeepMind and Anthropic, can command seven-figure salaries and are bombarded with job offers whenever they log on to LinkedIn. More senior engineers can earn up to $10 million.
“There is no question that there is an enormous talent war,” says Jordan Jacobs, managing partner of Toronto-based Radical Ventures, which has invested in some 50 AI start-ups globally. “At the top of the pyramid are the people who can build foundation models and make them sing. There are very few people who can do that effectively and there are organizations that will pay them an absolute fortune to do so.”
But it is also a further sign that the emerging AI economy will probably be dominated by the US tech giants with the money, human capital and cloud computing infrastructure needed to train state of the art foundation models, such as OpenAI’s GPT-4 and Google’s Gemini. (Source: ft.com, italics mine)
6. Re-posting this item from March 9th:
When it comes to the artificial intelligence that powers chatbots like ChatGPT, China lags behind the United States. But when it comes to producing the scientists behind a new generation of humanoid technologies, China is pulling ahead.
New research shows that China has by some metrics eclipsed the United States as the biggest producer of A.I. talent, with the country generating almost half the world’s top A.I. researchers. By contrast, about 18 percent come from U.S. undergraduate institutions, according to the study, from MacroPolo, a think tank run by the Paulson Institute, which promotes constructive ties between the United States and China.
The findings show a jump for China, which produced about one-third of the world’s top talent three years earlier. The United States, by contrast, remained mostly the same. The research is based on the backgrounds of researchers whose papers were published at 2022’s Conference on Neural Information Processing Systems. NeurIPS, as it is known, is focused on advances in neural networks, which have anchored recent developments in generative A.I. (Sources: news-items.substack.com, nytimes.com, macropolo.org)
7. When thousands of security researchers descend on Las Vegas every August for what's come to be known as “hacker summer camp,” the back-to-back Black Hat and Defcon hacker conferences, it's a given that some of them will experiment with hacking the infrastructure of Vegas itself, the city's elaborate array of casino and hospitality technology. But at one private event in 2022, a select group of researchers were actually invited to hack a Vegas hotel room, competing in a suite crowded with their laptops and cans of Red Bull to find digital vulnerabilities in every one of the room's gadgets, from its TV to its bedside VoIP phone. One team of hackers spent those days focused on the lock on the room's door, perhaps its most sensitive piece of technology of all. Now, more than a year and a half later, they're finally bringing to light the results of that work: a technique they discovered that would allow an intruder to open any of millions of hotel rooms worldwide in seconds, with just two taps. (Source: wired.com)
8. WSJ Pro:
A data- and cybersecurity startup has raised $60 million in a growth round, part of a wave of funding for companies that protect the information that gets fed into generative AI systems.
New York-based BigID said the round, led by Riverwood Capital with participation by Silver Lake Waterman and Advent International, values the company at over $1 billion. BigID generates nearly $100 million in yearly revenue, the company said.
Captivated by the promise of generative AI, companies are beginning to use it to leverage their massive data repositories to increase productivity and unlock new business lines. But that data is at higher risk of being hacked and manipulated—an encroachment known as data poisoning—investors and entrepreneurs say.
Venture funding globally for startups offering cybersecurity for artificial intelligence applications, including large language models, jumped 81% last year to $224 million, according to analytics firm PitchBook Data. Behind the surge is a broad belief that security must move in lockstep with AI advancements. (Source: wsj.com)
9. Paul Kedrosky:
I regularly speak with some of the largest investors in the world about the shifting economic and technology ground on which we stand, as waves of AI make it all seem less permanent than it once did. One of the questions I get a lot—and I got it this week in a private meeting with some executives at a major U.S. bank—is, "Who is the next Microsoft or Google or whatever?" They want to know, in short, who wins?
My answer, in general, is that they are thinking about it all wrong. In its current form, AI is job replacement technology, something that will remove humans from the workplace, not augment them. Economist Daron Acemoglu at MIT, among others, has written extensively about this, about augmenting vs replacing technologies, and he is worth reading on the topic……
What I generally tell people is that the winners—and this is the wrong and sociopathic word—will be dominated by companies using these technologies to slash costs. We will see firms in people-heavy but rule-bound fields, like insurance, airline booking, call centers, and so on, discover they can run with a fraction of their current employees. Shareholders will reward this as margins rise, competitors and boards will take notice, and every other company will follow, quickly shedding costs (read: people) and driving up margins.
Consumers will benefit from this, of course, in the same way they benefited from the rise of Costco and Walmart. Efficiencies, consolidation, and lower prices have diffuse and immense benefits, even if they have concentrated costs. The next Microsoft is a diffuse and fast-moving anti-Microsoft of cost-cutting. (Source: paulkedrosky.com)
10. China has 12½ times more robots in its workforce than expected, according to an independent think tank in Washington, an indication that Chinese workers are being replaced by robots at the fastest rate in the world. “China does not yet appear to be leading in robotic innovation, but … it is likely only a matter of time before Chinese robotics companies catch up to the leading edge,” the Information Technology and Innovation Foundation (ITIF) said in a report released on Monday. The decision to replace a human worker with a robot is often about saving money. It follows that developed countries with high wages would have greater penetration rates of robots than do lower-wage countries. But, the ITIF found that China is using automation far more than expected based on what workers get paid in manufacturing, with 12½ times more robots in use than predicted – a big jump from 1.6 times more in 2017. In contrast, the United States uses only 70 per cent of the robots it should be using, given the wages paid to its manufacturing workers. (Source: scmp.com, itif.org)
11. Apple chief Tim Cook has said China is “critical” to its business in comments made during a visit to the country, where consumers are showing signs of souring on the company’s iPhones and US technology. Amid rising geopolitical tensions, Cook has become a frequent visitor to China, a key market and huge manufacturing base for Apple products. His latest public relations offensive kicked off in Shanghai with an interview with local media, in which he delivered effusive praise and a promise to invest more in the country. The state-owned Global Times, best known for its nationalist bashing of the US, reported the increased investment pledge and quoted Cook as saying: “There’s no supply chain in the world that’s more critical to us than China.” Shanghai state-owned media noted Cook praised the “high level of modernisation in Chinese factories, with very advanced manufacturing capabilities and well-trained workers”. (Source: ft.com)
12. From yesterday’s New York Times:
General Motors said Friday that it had stopped sharing details about how people drove its cars with two data brokers that created risk profiles for the insurance industry.
The decision followed a New York Times report this month that G.M. had, for years, been sharing data about drivers’ mileage, braking, acceleration and speed with the insurance industry. The drivers were enrolled — some unknowingly, they said — in OnStar Smart Driver, a feature in G.M.’s internet-connected cars that collected data about how the car had been driven and promised feedback and digital badges for good driving.
Some drivers said their insurance rates had increased as a result of the captured data, which G.M. shared with two brokers, LexisNexis Risk Solutions and Verisk. The firms then sold the data to insurance companies.
Since Wednesday, “OnStar Smart Driver customer data is no longer being shared with LexisNexis or Verisk,” a G.M. spokeswoman, Malorie Lucich, said in an emailed statement. “Customer trust is a priority for us, and we are actively evaluating our privacy processes and policies.” (Source: nytimes.com, italics mine)
13. A man in Denmark was sentenced to 18 months in prison yesterday for using fake accounts to trick music streaming services into paying him 2 million Danish kroner ($290,000) in royalties. The unusual case reveals a weak spot in the business model behind the world’s biggest music platforms. The 53-year-old consultant, who had pleaded not guilty, was convicted of data fraud and copyright infringement after using bots to listen to his own music through fake profiles on both Spotify and Apple Music, collecting royalties in the process. The data fraud took place between 2013 and 2019. Fake or “artificial” streams are a big problem for the streaming industry. Between 1 billion and 3 billion fake streams took place on popular music platforms in 2021, according to a study by France’s National Music Center. Fake streams are a problem, according to the music industry, because they divert royalty payments away from real artists and pollute streaming platforms’ data. (Source: wired.com)
14. Patrick Honner:
Say you’re at a party with nine other people and everyone shakes everyone else’s hand exactly once. How many handshakes take place? This is the “handshake problem,” and it’s one of my favorites. As a math teacher, I love it because there are so many different ways you can arrive at the solution, and the diversity and interconnectedness of those strategies beautifully illustrate the power of creative thinking in math. One solution goes like this: Start with each person shaking every other person’s hand. Ten people, with nine handshakes each, produce 9 × 10 = 90 total handshakes. But this counts every handshake twice — once from each shaker’s perspective — so the actual number of handshakes is 90/2 = 45. A simple and lovely counting argument for the win! There’s also a completely different way to solve the problem. Imagine that the guests arrive one at a time, and when they get there, they shake hands with everyone present. (Source: quantamagazine.org)
15. Major League Baseball announced Friday it has opened an investigation into the allegations surrounding Dodgers star Shohei Ohtani and his now ex-interpreter Ippei Mizuhara. “Major League Baseball has been gathering information since we learned about the allegations involving Shohei Ohtani and Ippei Mizuhara from the news media. Earlier today, our Department of Investigations (DOI) began their formal process investigating the matter.” The investigation comes two days after The Times first reported that representatives of Ohtani accused Mizuhara, his longtime interpreter and close friend, of engaging in a “massive theft” of the two-way star’s funds, with millions of dollars of Ohtani’s money allegedly used to pay off gambling debts Mizuhara owed to an illegal bookmaker. (Source: latimes.com)
16. Boola-Boola. Yale defeats Auburn. (cbssports.com)
17. The 10 AI companies to watch right now. (Source: bloomberg.com)