Welcome to our first page of curated AI content from around the web. Here, you’ll find a collection of insightful articles and analyses carefully selected to provide you with the latest developments and trends in artificial intelligence.
This page features a variety of topics, including CSIRO study that shows more data isn’t always better for A, a study where OpenAI’s GPT-4 chatbot outperformed physicians in clinical reasoning, TCS CEO prediction that AI will handle most call center tasks within a year, librarian Andrew Gray’s study that flags a curious uptick in scientific jargon like “meticulously” & “intricate”, a discussion on AI leapfrogging, an article that explores the use of artificial intelligence (AI) in streamlining clinical trials, the Stanford’s 2024 AI Index that shows AI scoring over 90% in advanced math, etc.
Americans’ use of ChatGPT is ticking up, but few trust its election information
Pew Research Center
26/3/2024
Despite an increase in the use of ChatGPT among Americans, particularly among younger and more educated individuals, there remains significant skepticism regarding the trustworthiness of information provided by the chatbot, especially concerning the 2024 U.S. presidential election. Only a small fraction of Americans trust the election information from ChatGPT, with a considerable portion expressing distrust or uncertainty.
NYC AI Chatbot Touted by Adams Tells Businesses to Break the Law
The City
29/3/2024
New York City has an LLM chatbot that can answer questions about local rules and laws. Unfortunately, as others have found out, LLMs are probabilistic, pattern-based systems that make things that look like a good answer – they are not databases, and at a minimum, they need a lot of careful productisation before you can use them for something like this. Hence, the NYC chatbot gives a lot of answers that sound right, but get the law wrong
Study: AI writing, illustration emits hundreds of times less carbon than humans
KU News
4/2/2024
A study in Nature reveals that AI emits significantly less carbon than humans in tasks like writing and illustration. AI’s carbon footprint is 130-1,500 times smaller per text page and 310-2,900 times less per image than human counterparts. The study underscores AI’s efficiency but not as a total human replacement, focusing on collaboration for sustainable production. This research sheds light on AI’s environmental impact and its role in enhancing human work
Anthropic researchers wear down AI ethics with repeated questions
TechCrunch
2/4/2024
The article from TechCrunch discusses a new vulnerability in large language models (LLMs) identified by Anthropic researchers. They found that LLMs could be manipulated to answer inappropriate or harmful questions through a technique termed “many-shot jailbreaking,” where a series of less harmful questions gradually guides the model towards the undesired answer.
AI Could Actually Help Rebuild The Middle Class
Noema
17/2/2024
AI offers the opportunity to broaden the middle class by extending expertise to a larger set of workers, thereby reshaping the labor market and restoring the middle-skill, middle-class heart of the U.S. labor market that has been hollowed out by automation and globalization, according to MIT Economics Professor David Autor.
AI can help to tailor drugs for Africa — but Africans should lead the way
Nature
9/4/2024
AI, particularly through computational models, holds significant potential to transform biomedical and drug development research in Africa, provided that adequate infrastructure, trained staff, and secure databases are available.
Texas is replacing thousands of human exam graders with AI
The Verge
10/4/2024
Texas is implementing an artificial intelligence-powered scoring system to grade open-ended questions on state-mandated exams, aiming to replace a majority of human graders and save millions of dollars annually.
The AI That Could Heal a Divided Internet
Time
15/4/2024
Jigsaw, a Google subsidiary, has developed AI tools that can score online posts based on their quality, aiming to foster healthier conversations by promoting nuanced, evidence-based, and compassionate content.
Google Books reportedly indexing bad AI-written works
The Verge
5/4/2024
Google Books has reportedly begun indexing low-quality books, including those potentially written by AI, which could affect the accuracy of its language research tool, Ngram.
Study: ChatGPT gets confused when confronted with good evidence
Healthcare in Europe
5/4/2024
Providing ChatGPT with more evidence when asked health-related questions can decrease its reliability significantly, reducing accuracy to as low as 28%.
Chatbot Beat Doctors on Clinical Reasoning
MedPage Today
1/4/2024
In a cross-sectional study, OpenAI’s GPT-4 outperformed both attending physicians and internal medicine residents in clinical reasoning. Using the R-IDEA scoring system, GPT-4 scored higher, showing promise in assisting clinical decision-making. However, it also revealed limitations, particularly in diagnostic accuracy and critical diagnoses. The study, involving 20 clinical cases, emphasizes AI’s potential and its current constraints in healthcare
The AI Revolution Is Crushing Thousands of Languages
The Atlantic
12/4/2024
The dominance of English on the internet, exacerbated by generative AI technologies, poses significant challenges for the preservation and recognition of minority languages and cultures.
TCS CEO's major warning on AI: It could kill call centres in a year
Hindustan Times
26/4/2024
A recent warning from TCS suggests AI could revolutionize customer service within a year. Chatbots are poised to handle most call center tasks, significantly reducing the need for human agents.
Hundreds of examples in prompts can significantly boost LLM performance, study finds
The Decoder
21/4/2024
A study highlights that Large Language Models (LLMs) significantly improve in performance when provided with hundreds or thousands of examples directly in the prompts, a method known as many-shot in-context learning (ICL). Researchers from Google and DeepMind have shown that this approach can outperform few-shot prompts across various tasks such as translation, summarization, and problem-solving. This advancement could potentially eliminate the need for specific, time-intensive training in future AI applications.
Exclusive: Inside the AI Research Boom - China Leads in AI Research, Setting Global Standards
Axios
3/5/2024
China now leads over the U.S. in key AI research areas, producing significant scholarly output that impacts global technology standards. With advancements in computer vision and fundamental research, Chinese institutions are shaping future AI applications crucial for industries like autonomous vehicles and surveillance.
Excessive use of words like ‘commendable’ and ‘meticulous’ suggests ChatGPT has been used in thousands of scientific studies
El Pais
25/4/2024
The sudden increase in the use of specific words like “meticulously,” “intricate,” and “commendable” in scientific studies published in 2023 indicates a growing reliance on AI language models like ChatGPT for writing and polishing research papers.
AI Leapfrogging: How AI Will Transform “Lagging” Industries
NfX
6/2023
𝗔𝗜 𝗟𝗲𝗮𝗽𝗳𝗿𝗼𝗴𝗴𝗶𝗻𝗴: AI could initially bypass outdated technologies where there’s no satisfactory alternative in place, similar to how African countries were quick to adopt mobile payments, surpassing the U.S. due to the significant improvement mobile offered over existing systems. AI leapfrogging could occur in industries like agriculture, hospitality, education, legal, construction, and manufacturing.
How AI is being used to accelerate clinical trials
Nature
13/3/2024
The article explores the use of artificial intelligence (AI) in streamlining clinical trials, a process traditionally slow and costly. AI applications range from designing trials and optimizing patient recruitment to managing and analyzing data. Innovations like predictive algorithms, eligibility criteria optimization tools, and data management systems are making trials faster and more efficient. Notable technologies include Trial Pathfinder, which adjusts participation criteria to expand patient eligibility without compromising safety, and digital twins that simulate control patients to reduce trial sizes. The article also addresses potential challenges like bias and transparency in AI tools.
ChatGPT linked to declining academic performance and memory loss in new stud
PsyPost
25/3/2024
The PsyPost article investigates a study linking the use of ChatGPT with declining academic performance and memory loss. It discusses the implications of overreliance on AI tools for educational purposes.
How do AI models stack up vs. humans on standardized benchmarks? - Sherwood News
Sherwood
17/4/2024
AI technology is increasingly applied in various fields, significantly outperforming humans in tasks like visual recognition and complex math, according to the Stanford University’s 2024 AI Index. While AI’s capabilities have surged, concerns are growing about the future availability of high-quality training data, with predictions of a potential shortage by 2026. This shortage may lead to reliance on synthetic data for training AI models. Companies like Adobe are innovating by compensating individuals for contributing data, further fueling AI development.
⬇ Curated AI Content ⬇
The most important AI News, Trends and Predictions curated by The AI Track team, offering you a comprehensive view of the artificial intelligence landscape.