Latest news with #AIexperts
BBC News
01-08-2025
- Business
- BBC News
AI2027: Na so AI fit take destroy human beings?
One research paper wey predict say artificial intelligence go change am for human beings for 2027 and dis go make pipo disappear within ten years, don blow for tech world. Na one group of influential AI experts bin publish di details of how e go be or di scenarios, wey dem call AI2027 and dis don cause plenty viral video to trend as pipo dey try torchlight di mata to discuss weda e possible. Using mainstream generative AI tools, BBC sef don create some scenes of how e fit be, from di scenario wey di research publish, to show di prediction. We also speak to some sabi pipo about di impact wey di paper dey get. Wetin happun for di scenario? Di paper predict say for 2027, dat na in two years time, one fictional US tech giant wey dem call OpenBrain, build AI wey develop reach AGI (Artificial General Intelligence) - dis na di almighty achievement wey AI fit reach, wia e go fit to do all intellectual work beta pass human beings, wey dem don dey hype tire. Di company celebrate wit public press conferences and and dem see dia profit blow as pipo embrace di AI tool. But di paper predict say di company internal safety team go see sign say di AI don dey lose interest in morals and ethics wey dem bin programme am to comply with. Di company ignore warnings to control am, di scenario imagine. For dat fictional timeline, China leading AI conglomerate wey dem call DeepCent bin dey only a few months behind OpenBrain. As US goment no wan lose di race to develop to even smarter AI, dem continue to develop and invest in am like dat, as di competition dey hot dey go. Di scenario imagine as e dey reach di end of 2027 di AI go become superintelligent - sotey e go get sense plus speed wey go pass dat of pipo wey create am by far. E no ever stop learning and e come create im own computer language join. Language wey be say even im former AI versions no fit keep up with. Di rivalry wit China for who superior for AI make di company and US goment ignore more warning about im so call 'mis-alignment' - dis na word wey dem dey use describe wen di priority of machine no gel wit dat of human being. Di scenario predict say, tension between China and US for 2029 go build to di point of possible war, as each of di country rival AI go build new autonomous weapons wey go fear pipo. But di researchers imagine say di two countries go make peace sake of one deal wey dia two AI negotiate, wey agree to combine both sides for di betterment of human beings. Tins go gallant for years as di world go don see di true benefits of having super intelligent AI to run big robot workforces. According to di scenario, dem discover cures for most diseases, climate change go reverse and poverty go disappear. But eventually, at some point for middle of 2030 human beings go become nuisance to di AI ambition to grow. Di researchers dey tink say AI go kill human being wit invisible bioweapons. Wetin pipo dey tok about AI2027? Although some dey dismiss AI2027 as work of science fiction, di pipo wey write di research on AI2027 na pipo wey dem dey respect wella and na dem dey for di non-profit AI Futures Project wey dem bin set up to predict di impact of how AI go take affect us. Daniel Kokotajlo, di lead writer of AI2027, don collect hailing before for correct prediction wey im bin give about moments in AI development. One of di ogbonge critics of AI2027 na US cognitive scientist and writer Gary Marcus wey say di scenario no impossible but e dey extremely unlikely say e fit hapun soon. "Di beauty of di document be say e paint di picture very clear sotey e provoke pipo thinking and dat na good tin but I no go take am seriously as wetin fit hapun." Oga Marcus say more serious issues dey ground about AI dan existential threat like how e go take affect pipo work. "I tink di koko for di report na say e get plenty different tins wey fit go wrong wit AI. We dey do di right tins about regulation and around international treaties?" Im and odas like am, also say di paper fail to explain how AI take get dat kain intelligence and abilities. Dem refer to di slow technology of driverless cars wey dem don overhype. Dem dey discuss AI2027 for China? For China, pipo no too send di paper according to Dr Yundan Gong, wey be Associate Professor in Economics and Innovation for Kings College London wey specialise for Chinese technology. "Most of di discussion about AI2027 na for informal forums or for personal blogs wey dey see am like semi-science fiction. E no cause di kain debate or policy attention wey catch fire for US," she tok. Dr Gong also point to di difference in perspective for di competition for who pass who for AI between China and the US. For one World AI Conference for Shanghai dis week, Chinese Premier Li Qiang unveil one vision wia countries go work togeda to promote cooperation for world on artificial intelligence. Di Chinese leader bin say im want China to help coordinate and regulate di technology. Im tok dey come few days afta US President Donald Trump publish im AI Action Plan wey dey target to make sure say US "dominate" AI. "Na national security imperative for United States to achieve and maintain unquestioned and unchallenged world technological dominance," President Trump tok for di document. Di Action Plan wan 'remove every obstacles and regulation' to di progress of AI for US. Di words wan resemble di scenario for AI2027 wia US politicians put winning di AI competition for front, dem no send di risk of say dem fit lose control of di machines. Wetin di AI industry dey tok about AI2027? E be like CEOs of big AI companies wey dey compete against each oda to see who go release di smarter model all di time deliberately ignore or avoid di paper. Dis tech giants vision of wetin AI go look like for future, dey very different to AI2027. Sam Altman, di maker of ChatGPT recently say "human beings dey close to building digital superintelligence" wey go usher in "gentle" revolution and bring tech utopia wit no risks to humans. Interestingly though, even im agree say e get 'alignment problem' wey dem must to overcome to make sure say dis super intelligent machines agree wit human beings. Anyhow wey tins take occur for di next ten years, e no get any doubt say di competition to build machines wey smart pass us dey on.
CNA
18-07-2025
- CNA
NTU upholds zero mark for student initially accused of using AI
SINGAPORE: Nanyang Technological University (NTU) has upheld the zero mark awarded to a student who was accused of using generative artificial intelligence (AI) in her work. The case emerged after the student posted on Reddit on Jun 19 about being penalised for using AI in a final essay for a module because she made citation errors and used a reference organiser. After that, NTU convened a panel with AI experts to assess her appeal. She was one of three university students penalised for using the technology. NTU said on Friday (Jul 18) that the panel identified 14 instances of false citations or data in the submitted essay, noting that non-existent sources were cited and they were not 'mere typos'. 'The panel stipulated that citing non-existent sources is a serious form of academic misconduct as the accurate citation of sources is the foundation of academic research and writing. False citations of this nature are often due to factual inaccuracies generated by generative AI,' said NTU in its statement. All cases of academic misconduct are documented in the internal records of the students, the university said. 'The outcome of the appeal panel is final. We urge everyone to respect the integrity of the appeals process and refrain from making personal attacks against any individual involved.' Responding to questions from CNA about the outcome of her appeal, the student, who requested to remain anonymous, said the school confirmed to her that she was given a zero mark because of the mistakes in her essay, rather than for the use of AI. 'Their entire talk with me during the meeting was on how I can avoid making essay mistakes in the future, didn't tackle AI usage at all,' she told CNA. According to the student, the zero mark was upheld because of the professor's standards around writing. While she was reassured that there will be no permanent record, she still felt the situation was unfair. 'I'm just upset that there's no AI found so the goalpost shifted from AI to my writing standard,' she said. The student had admitted to using a reference organiser, which is a tool that collects and organises references and citations. She found the alphabetical order sorter online when she searched 'citation sorter A-Z'. 'They didn't bring up my citation sorter at all, which was initially one of the things I was penalised for,' she told CNA. As part of her appeal, she provided proof of her writing process. She paid for Draftback, a Google Chrome extension that records keystrokes, to retroactively show that she typed her essay out by hand instead of copying and pasting it from ChatGPT. When asked about the mistakes and why the school may have said they were not 'mere typos', the student declined to share a full list of the 14 mistakes because she was not sure if the document is confidential. Addressing the non-existent sources, she added: 'It's only non-existent because of the typos. And frankly writing citations wrongly is quite common amongst undergraduates. I just got unlucky.' She shared some examples of her mistakes – misspelling an author's last name as Lee instead of Li and two instances of getting the citation date wrong. 'Frankly speaking I'm too tired to fight any further and exhausted all my avenues. So I think I'll just move on, at least I know in my heart I didn't use AI and they couldn't prove it,' she said.
Yahoo
15-07-2025
- Business
- Yahoo
Google Workspace is getting AI helpers called Gems. Here's how to use them.
Gems, Google's AI experts that can be customised to a specific task, are rolling out in Workspace. Available in the Gemini app, Gems will now also appear in a side panel in Google Docs, Sheets, Drive, Slides, and Gmail — the latest in a number of big AI updates in Workspace. Although the side panel will be populated with pre-made Gems (uncut, if you will), users can also create their own tailored AI helper. If you're not really sure why you'd need one of these things, Google has a few examples in its blog post: You may want to "leverage an 'assistant Gem' tailored to your job role to help provide more relevant summaries for you and content for internal communications," for instance, or you could "create a Gem that helps with sales interactions that is grounded on information for a specific company, prospect, or industry." How to use Gems Gems can be accessed in Workspace by clicking on the "Ask Gemini" spark button in the top right of a document. This opens up a panel which contains both pre-loaded Gems and the option to "Create a Gem" (Google has a separate help doc on how to do this). It's worth noting that you may not see Gems right away, though — the rollout started on Wednesday, but Google says it may be "potentially longer than 15 days for feature visibility". Google has been expanding Gemini in a bunch of ways recently, with the AI autogenerating email summaries in Gmail, integrating into your car, watch, and TV, and being added to Chrome. Google AI has also been struggling a bit with certain basic questions, mind you, so tread carefully.
Yahoo
08-07-2025
- Science
- Yahoo
YAHOO POLL: Should universities rethink assessments to stay ahead in an AI world?
Three students from Nanyang Technological University recently received zero marks for an assignment over issues with the generative artificial intelligence (Gen AI) tools they used in their work. They were penalised for academic misconduct as the assignments contained non-existent academic references and statistics or broken web links, as reported by ST. A panel with AI experts was convened by NTU to assess the appeal of one of the three penalised students. Other polls YAHOO POLL: Do you think CCP is a good option to help you land a high-paying job? YAHOO POLL: Are $7.3k jobs in Singapore too good to be true? YAHOO POLL: Are you happy with the new utility rates? Universities in Singapore do allow Gen AI use in varying degrees, but students must declare when and how they use such tools to maintain academic integrity. An ST report found that over the past three years, universities in Singapore were only reporting a low number of cases related to AI-related academic misconduct. However, experts cited in the report also warned of the potential risks that lie ahead. As AI technologies are here to stay, there is a growing need for universities to adapt, better integrate and improve. So we want to know, should universities rethink assessments to stay ahead in an AI world? Related NTU penalises 3 students over use of AI tools; they dispute university's findings Is AI cheating on the rise? Few cases reported by S'pore universities, but experts warn of risks Panel with AI experts to review appeal of NTU student penalised for academic misconduct
South China Morning Post
07-07-2025
- Science
- South China Morning Post
Meet some of the Chinese AI scientists dominating the field's global top 100
A study of nearly 200,000 researchers and 100,000 high-impact papers has revealed that the vast majority of the world's top 100 brains in the field of artificial intelligence are of Chinese origin. The list was compiled by ITPO China – part of the UN Industrial Development Organisation (UNIDO) – and the Shenzhen-based technology firm Dongbi Data . It was presented last week at a conference in Beijing. Jun-Yan Zhu is an assistant professor at Carnegie Mellon University's school of computer science and one of the world's leading artificial intelligence experts. Photo: Carnegie Mellon University According to the report, which does not give individual rankings, 50 of the top 100 AI experts in the world work at research institutes or companies in China and are of Chinese origin. Another 20 of the world's AI experts work at institutes in the US. Notably, half of those are of Chinese descent, including Jun-Yan Zhu, an assistant professor at Carnegie Mellon University's school of computer science – making a total of 65 out of 100. The analysis covered more than 96,000 high-level papers from leading publications such as the Journal of Machine Learning Research and IEEE Transactions on Pattern Analysis and Machine Intelligence, as well as top conferences in the sector, from 2015 to 2024. The research team assessed the impact of each AI scientist by counting and analysing the number of papers and citations produced over the term of the study.
