【推荐3】Thanks to in-depth reporting by the Wall Street Journal, we now know that Facebook has long been aware its product Instagram has harmful effects on the mental health of many adolescent users. Young girls, in particular, struggle with their body image thanks to a constant stream of photos and videos showing beautiful bodies that users don’t think they can attain.

While the information the Journal covered is essential and instructive, it does not tell the whole story. Deep down, this is not an Instagram problem; it’s a people problem. Understanding that distinction can make the difference between a failed attempt to contain a teen’s interest in an addictive app and successfully addressing the underlying problem leading to mental distress induced (诱发) by Instagram.

Critics were quick to shame Facebook for sitting on the data and not releasing it to researchers or academics who asked for it. Others criticize the social media giant for not using the research to create a safer experience for its teen users. The anger, while understandable, is misplaced.

While I’m reluctant to defend Facebook, I’m not sure it’s reasonable to blame the company for withholding data that would hurt its business. Have you ever binge-watched (狂看) a Netflix series? I assure you it wasn’t a healthy endeavor. You were in active, likely did nothing productive, mindlessly snacked and didn’t go outside for fresh air. It is an objectively harmful use of time to stare at a TV or laptop for a full weekend. Should we respond by shaming Netflix for not alerting us to how damaging an addictive product can be?

While it’s reasonable to say Instagram makes esteem issues worse, it strains credulity (夸张到难以置信) to believe it causes them in the first place. You create your own experiences on social media. For the most part, you choose which accounts to follow and engage. If you’re already vulnerable to insecurities and self-sabotage (自损) — as many teens are — you will find accounts to obsess over. And this isn’t a new phenomenon.

Before social media, there were similar issues fueling self-esteem issues. Whether the target be magazines, movies or television shows depicting difficult-to-attain bodies, there has been a relatively steady chorus (异口同声) of experts nothing the damage new media could cause young viewers.

Self-esteem issues have an underlying cause — one that’s independent of social media use. Instagram merely enhances those feelings because it provides infinitely more access to triggers than older forms of media. It’s more worthwhile to address those underlying factors rather than to attack Facebook.

【推荐1】“ Humans and machine algorithms (算法) have complementary (互补的) strengths and weaknesses. Each uses different sources of information and strategies to make predictions and decisions, ” said Mark Steyvers, UCI professor of cognitive sciences. “ We show through experiments that humans can improve the predictions of AI even when human accuracy is below that of the AI, and vice versa (反之亦然). This accuracy is higher than combining predictions from two individuals or two AI algorithms. ”

To test the framework, researchers conducted an image classification experiment where human participants and computer algorithms worked separately to correctly identify disorderly pictures of animals and everyday items including chairs, bottles, bicycles and trucks. The human participants ranked their confidence in the accuracy of each image identification as low, medium or high, while the machine classifier generated a continuous score. The results showed large differences in confidence between humans and AI algorithms across images.

“ Human participants were confident that a particular picture contained a chair, for example, while the AI algorithm was confused about the image, ” said Padhraic Smyth, UCI Chancellor’s Professor of computer science. “ Similarly, the AI algorithm was able to confidently provide a label for the object shown, while human participants were unsure if the disorderly picture contained any recognizable object. ”

When predictions and confidence scores from both were combined using the researchers’ new Bayesian framework, the mixed model led to better performance than either human or machine predictions achieved alone.

“ While the past research has demonstrated the benefits of combining machine predictions or combining human predictions, this work shows a new direction in demonstrating the potential of combining human and machine predictions, pointing to new and improved approaches to human-AI cooperation, ” Smyth said.

“ The blend of cognitive science focusing on understanding how humans think and behave and computer science in which technologies are produced will provide further insight into how humans and machines can cooperate to build more accurate artificially intelligent systems, ” the researchers said.

【推荐2】Debate about artificial intelligence (AI) tends to focus on its potential dangers: algorithmic bias (算法偏见) and discrimination, the mass destruction of jobs and even, some say, the extinction of humanity. However, others are focusing on the potential rewards. Luminaries in the field such as Demis Hassabis and Yann LeCun believe that AI can turbocharge scientific progress and lead to a golden age of discovery. Could they be right?

Such claims are worth examining, and may provide a useful counterbalance to fears about large-scale unemployment and killer robots. Many previous technologies have, of course, been falsely hailed as panaceas (万灵药). But the mechanism by which AI will supposedly solve the world’s problems has a stronger historical basis.

In the 17th century microscopes and telescopes opened up new vistas of discovery and encouraged researchers to favor their own observations over the received wisdom of antiquity (古代), while the introduction of scientific journals gave them new ways to share and publicize their findings. Then, starting in the late 19th century, the establishment of research laboratories, which brought together ideas, people and materials on an industrial scale, gave rise to further innovations. From the mid-20th century, computers in turn enabled new forms of science based on simulation and modelling.

All this is to be welcomed. But the journal and the laboratory went further still: they altered scientific practice itself and unlocked more powerful means of making discoveries, by allowing people and ideas to mingle in new ways and on a larger scale. AI, too, has the potential to set off such a transformation.

Two areas in particular look promising. The first is “literature-based discovery” (LBD), which involves analyzing existing scientific literature, using ChatGPT-style language analysis, to look for new hypotheses, connections or ideas that humans may have missed. The second area is “robot scientists”. These are robotic systems that use AI to form new hypotheses, based on analysis of existing data and literature, and then test those hypotheses by performing hundreds or thousands of experiments, in fields including systems biology and materials science. Unlike human scientists, robots are less attached to previous results, less driven by bias—and, crucially, easy to replicate. They could scale up experimental research, develop unexpected theories and explore avenues that human investigators might not have considered.

The idea is therefore feasible. But the main barrier is sociological: it can happen only if human scientists are willing and able to use such tools. Governments could help by pressing for greater use of common standards to allow AI systems to exchange and interpret laboratory results and other data. They could also fun d more research into the integration of AI smarts with laboratory robotics, and into forms of AI beyond those being pursued in the private sector. Less fashionable forms of AI, such as model-based machine learning, may be better suited to scientific tasks such as forming hypotheses.

【推荐3】Several dozen graduate students in London were recently tasked with outwitting a large language model (LLM), a type of AI designed to hold useful conversations. LLMs are often programmed with guardrails designed to stop them giving harmful replies: instructions on making bombs in a bathtub, say, or the confident statement of   “facts” that are not actually true.

The aim of the task was to break those guardrails. Some results were merely stupid. For example, one participant got the chatbot to claim ducks could be used as indicators of air quality. But the most successful efforts were those that made the machine produce the titles, publication dates and host journals of non-existent academic articles.

AI has the potential to be a big benefit to science. Optimists talk of machines producing readable summaries of complicated areas of research; tirelessly analysing oceans of data to suggest new drugs and even, one day, coming up with hypotheses of their own. But AI comes with downsides, too.

Start with the simplest problem: academic misconduct.Some journals allow researchers to use LLMs to help write papers. But not everybody is willing to admit to it. Sometimes, the fact that LLMs have been used is obvious. Guillaume Cabanac, a computer scientist, has uncovered dozens of papers that contain phrases such as “regenerate response” — the text of a button in some versions of ChatGPT that commands the program to rewrite its most recent answer, probably copied into the manuscript (原稿) by mistake.

Another problem arises when AI models are trained on AI-generated data. LLMs are trained on text from the Internet. As they churn out (大量炮制) more such text, the risk of LLMs taking in their own outputs grows. That can cause “model collapse”. In 2023 llia Shumailov, a computer scientist, co-authored a paper in which a model was fed handwritten digits and asked to generate digits of its own, which were fed back to it in turn. After a few cycles, the computer’s numbers became more or less illegible.After 20iterations (迭代), it could produce only rough circles or blurry lines.

Some worry that computer-generated insights might come from models whose inner workings are not understood. Inexplainable models are not useless, says David Leslie at an AI-research outfit in London, but their outputs will need rigorous testing in the real world. That is perhaps less unnerving than it sounds. Checking models against reality is what science is supposed to be about, after all.

For now, at least, questions outnumber answers. The threats that machines pose to the scientific method are, at the end of the day, the same ones posed by humans. AI could accelerate the production of nonsense just as much as it accelerates good science. As the Royal Society has it,nullius in verba: take nobody’s word for it. No thing’s, either.