The AI Scientist

Large language models, the AI algorithms taking the world by storm, are galvanizing academic research.

These algorithms scrape data from the digital world, learn patterns in the data, and use these patterns to complete a variety of specialized tasks. Some algorithms are already aiding research scientists. Some can solve challenging math problems. Others are “dreaming up” new proteins to tackle some of our worst health problems, including Alzheimer’s and cancer.

Although helpful, these only assist in the last stage of research—that is, when scientists already have ideas in mind. What about having an AI to guide a new idea in the first place?

AI can already help draft scientific articles, generate code, and search scientific literature. These steps are akin to when scientists first begin gathering knowledge and form ideas based on what they’ve learned.

Some of these ideas are highly creative, in the sense that they could lead to out-the-box theories and applications. But creativity is subjective. One way to gauge potential impact and other factors for research ideas is to call in a human judge, blinded to the experiment.

“The best way for us to contextualize such capabilities is to have a head-to-head comparison” between AI and human experts, study author Chenglei Si told Nature.

The team recruited over 100 computer scientists with expertise in natural language processing to come up with ideas, act as judges, or both. These experts are especially well-versed in how computers can communicate with people using everyday language. The team pitted 49 participants against a state-of-the-art LLM based on Anthropic’s Claude 3.5. The scientists earned $300 per idea plus an additional $1,000 if their idea scored in the top 5 overall.

The Human Critic

To make it a fair test, the judges didn’t know which responses were from AI. To disguise them, the team translated submissions from humans and AI into a generic tone using another LLM. The judges evaluated ideas on novelty, excitement, and—most importantly—if they could work.

After aggregating reviews, the team found that, on average, ideas generated by human experts were rated less exciting than those by AI, but more feasible. As the AI generated more ideas, however, it became less novel, increasingly generating duplicates. Digging through the AI’s nearly 4,000 ideas, the team found around 200 unique ones that warranted more exploration.

But many weren’t reliable. Part of the problem stems from the fact the AI made unrealistic assumptions. It hallucinated ideas that were “ungrounded and independent of the data” it was trained on, wrote the authors. The LLM generated ideas that sounded new and exciting but weren’t necessarily practical for AI research, often because of latency or hardware problems.

“Our results indeed indicated some feasibility trade-offs of AI ideas,” wrote the team.

Full Article

posted by f.sheikh

(Sarah Anjum Bari on the Power of Visible Words)

Amar bhai ke marli keno?: Why did you kill my brother?

Past midnight on August 11, as my parents drive me to Dhaka’s Shahjalal International Airport, the city I call home seems locked in a tense quiet. No one has gone to sleep, our WhatsApp groups and Facebook feeds remind us—it’s been a month since Bangladeshis went to bed at a decent hour—but the roads are empty of both pedestrians and traffic police. The few cars and trucks passing through are shepherded by cap-wearing, whistle-wielding students, the streets somehow more ordered in their untrained hands. Entrances to neighborhoods are being manned by volunteer groups from resident families. Their stake out is a precaution against a recent surge of burglaries.

In this absence of Dhaka’s usual nighttime soundscape, what calls out to us on the roads are the red and black words streaked across the city’s walls, its pillars, shop fronts, gates, along the beams of flyovers and even the trunks of trees. At varying heights and in varying directions, we read, among other phrases:

Noy dofa dabi: A nine-point demand

Ek dofa dabi: A one-point demand

Dhaka, a city of corpses. 

Sheikh Hasina is a murderer. Sheikh Hasina must step down.

36 July. No more lies. The dictator has fled.

We are free.

Since the beginning of June 2024, public university students in Bangladesh had been protesting for a reform of the job quota that unfairly benefitted freedom fighters’ descendants, thereby benefiting supporters or relatives of the ruling party who had fought in the war. In a national address to the protests, the (now former) Prime Minister Sheikh Hasina mockingly inquired whether the jobs should instead be allocated to “razakars.” By using that term, the gravest of insults for a Bangladeshi, she had compared university students asking for a fair shot at employment to collaborators who aided Pakistan’s genocide against Bangladesh during the ‘71 Liberation War. Students and their allies across the country were outraged. They poured onto the streets demanding both apology and reform, carrying slogans that now endure on the city walls.

Tumi ke, ami ke? Razakar, razakar. Ke bolechhe, ke bolechhe? Shoirachar, shoirachar!

“Who are you, who am I? The dictator says we are razakars!”

Chaite gelam odhikar, hoye gelam razakar.

“We came to ask for our rights, only to be labeled traitors.”

Full Article

posted by f.sheikh

After years of debate and contemplation, I’ve come to believe in a Christian God of limited abilities. Here’s why: I rejected Christianity at the age of 14, upsetting my grandmother by refusing to get confirmed in the Catholic faith of my upbringing. Partly it was intellectual issues: why would a loving and all-powerful God create a world with so much suffering? Partly it was ethical issues. It was a time when I was questioning my sexuality, and it seemed to me wrong not to allow a gay person to flourish through a loving relationship with a partner they are attracted to. But, most of all, Christianity just seemed very unspiritual. I got very little out of boring church services, and it seemed to be all about pleasing the old guy in the sky so you get to heaven. Science and philosophy seemed a more rational way to make sense of life, which ultimately led me to become a philosophy professor. I now think the evidence points towards a hypothesis that John Stuart Mill took seriously: a good God of limited abilities. This hypothesis is able to account both for the imperfections of our universe – in terms of God’s limited abilities – and for the things about our universe that are improbably good, such as fine-tuning and psycho-physical harmony. God would have liked to make intelligent life in an instant, or by breathing into the dust as we see depicted in Genesis. But the only way God was able to create life was by bringing into existence a universe with the right physics that would eventually evolve intelligent life. God made the best universe they could.

Full Article

posted by f.sheikh

“While protests continue erupting with remarkable frequency, they are also failing, at historic rates, to achieve protesters’ stated goals. As Time hailed the power of the protester, the rate at which mass protests succeeded in meeting their objectives was plummeting, from two in three during the early 2000s to just one in six by the early 2020s. Activists are now reaping less fruit from their labour, while many would-be activists never take the plunge in the first place because they reasonably doubt that their participation will make any difference. Why aren’t protesters winning like they once did, and what would make protests more effective?”

Some scholars pin declining protest success rates on social media, which allows huge crowds to assemble without building the organisational structures and strong networks necessary to effect meaningful change. Others blame dictators’ use of ‘smart repression’ techniques, including censorship, propaganda and misinformation. As the political scientist Kurt Weyland points out, counterrevolutionaries have historically held an advantage over revolutionaries because they are willing to bide their time, heed advisors, and do their research on which repressive methods have worked in the past. Revolutionaries, in contrast, tend to leap into action, sometimes miscalculating their odds of success and choosing misguided strategies.

Violence seldom pays. From 1900 to 2006, nonviolent resistance campaigns were more than twice as effective as violent ones (though even nonviolent campaigns have struggled to achieve their goals in more recent years). One reason why violence backfires is that it discourages new activists from joining or otherwise supporting a movement.

Cohesive demands are more persuasive than mixed demands.

Diverse coalitions signal that a movement is more than a radical fringe. 

Marginalised protesters influence lawmakers more than privileged protesters.

Most protests fail. What are activists doing right when they win? | Psyche Ideas

posted by F.Sheikh