Watch: AI-powered robot learns to play badminton against humans
An artificial intelligence -driven legged robot has been trained to play badminton against human opponents with impressive agility.
It has been developed by a Swiss -led team at ETH Zurich, who used reinforcement learning — a type of AI that improves decision-making by learning from repeated attempts.
The robot can follow the shuttlecock and hit it precisely in fast-paced games.
It accurately predicts shuttlecock trajectories, navigates the game area, and competes with human players.
Researcher Yuntao Ma believes it showcases AI's potential to drive legged robots in complex tasks, which could pave the way for future advancements in autonomous, intelligent systems, including humanoids.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Daily Mail
3 hours ago
- Daily Mail
Lawyers warned to stop using ChatGPT to argue lawsuits after AI programs 'made up fictitious cases'
Lawyers in England and Wales have been warned they could face 'severe sanctions' including potential criminal prosecution if they present false material generated by AI in court. The ruling, by one of Britain's most senior judges, comes on the back of a string of cases in which which artificially intelligence software has produced fictitious legal cases and completely invented quotes. The first case saw AI fabricate 'inaccurate and fictitious' material in a lawsuit brought against two banks, The New York Times reported. Meanwhile, the second involved a lawyer for a man suing his local council who was unable to explain the origin of the nonexistent precedents in his legal argument. While large language models (LLMs) like OpenAI 's ChatGPT and Google 's Gemini are capable of producing long accurate-sounding texts, they are technically only focused on producing a 'statistically plausible' reply. The programs are also prone to what researchers call 'hallucinations' - outputs that are misleading or lack any factual basis. AI Agent and Assistance platform Vectera has monitored the accuracy of AI chatbots since 2023 and found that the top programs hallucinate between 0.7 per cent and 2.2 per cent of the time - with others dramatically higher. However, those figures become astronomically higher when the chatbots are prompted to produce longer texts from scratch, with market leader OpenAI recently acknowledging that its flagship ChatGPT system hallucinates between 51 per cent and 79 per cent of the time if asked open-ended questions. While large language models (LLMs) like OpenAI's ChatGPT and Google's Gemini are capable of producing long accurate-sounding texts, they are technically only focused on producing a 'statistically plausible' reply - which can lead to them 'hallucinating' false information Dame Victoria Sharp, president of the King's Bench Division of the High Court, and Justice Jeremy Johnson KC, authored the new ruling. In it they say: 'The referrals arise out of the actual or suspected use by lawyers of generative artificial intelligence tools to produce written legal arguments or witness statements which are not then checked, so that false information (typically a fake citation or quotation) is put before the court. 'The facts of these cases raise concerns about the competence and conduct of the individual lawyers who have been referred to this court. 'They raise broader areas of concern however as to the adequacy of the training, supervision and regulation of those who practice before the courts, and as to the practical steps taken by those with responsibilities in those areas to ensure that lawyers who conduct litigation understand and comply with their professional and ethical responsibilities and their duties to the court.' The pair argued that existing guidance around AI was 'insufficient to address the misuse of artificial intelligence'. Judge Sharp wrote: 'There are serious implications for the administration of justice and public confidence in the justice system if artificial intelligence is misused,' While acknowledging that AI remained a 'powerful technology' with legitimate use cases, she nevertheless reiterated that the technology brought 'risks as well as opportunities.' In the first case cited in the judgment, a British man sought millions in damages from two banks. The court discovered that 18 out of 45 citations included in the legal arguments featured past cases that simply did not exist. Even in instances in which the cases did exist, often the quotations were inaccurate or did not support the legal argument being presented. The second case, which dates to May 2023, involved a man who was turned down for emergency accommodation from the local authority and ultimately became homeless. His legal team cited five past cases, which the opposing lawyers discovered simply did not exist - tipped off by the fact by the US spellings and formulaic prose style. Rapid improvements in AI systems means its use is becoming a global issue in the field of law, as the judicial sector figures out how to incorporate artificial intelligence into what is frequently a very traditional, rules-bound work environment. Earlier this year a New York lawyer faced disciplinary proceedings after being caught using ChatGPT for research and citing a none-existent case in a medical malpractice lawsuit. Attorney Jae Lee was referred to the grievance panel of the 2nd U.S. Circuit Court of Appeals in February 2025 after she cited a fabricated case about a Queens doctor botching an abortion in an appeal to revive her client's lawsuit. The case did not exist and had been conjured up by OpenAI's ChatGPT and the case was dismissed. The court ordered Lee to submit a copy of the cited decision after it was not able to find the case. She responded that she was 'unable to furnish a copy of the decision.' Lee said she had included a case 'suggested' by ChatGPT but that there was 'no bad faith, willfulness, or prejudice towards the opposing party or the judicial system' in doing so. The conduct 'falls well below the basic obligations of counsel,' a three-judge panel for the Manhattan-based appeals court wrote. In June two New York lawyers were fined $5,000 after they relied on fake research created by ChatGPT for a submission in an injury claim against Avianca airline. Judge Kevin Castel said attorneys Steven Schwartz and Peter LoDuca acted in bad faith by using the AI bot's submissions - some of which contained 'gibberish' - even after judicial orders questioned their authenticity.
The Independent
8 hours ago
- The Independent
DOGE used flawed AI tool to ‘munch' Veteran Affairs contracts, report claims
Employees in the Department of Government Efficiency reportedly used a flawed artificial intelligence model to determine the necessity of contracts in the Department of Veterans Affairs, resulting in hundreds of contracts, valued at millions of dollars, being canceled. Given only 30 days to implement President Donald Trump 's executive order directing DOGE to review government contracts and grants to ensure they align with the president's policies, an engineer in DOGE rushed to create an AI to assist in the task. Engineer Sahil Lavingia wrote code which told the AI to cancel, or in his words 'munch,' anything that wasn't 'directly supporting patient care' within the agency. However neither he, nor the model, required the knowledge to make those decisions. ' 'I'm sure mistakes were made,' he told ProPublica. Mistakes are always made.' One of the key problems was that the AI only reviewed the first 10,000 characters (roughly 2,500 words) of contracts to determine whether it was 'munchable' – Lavingia's term for if the task could be done by VA staffers rather than outsourcing, ProPublica reported. Experts who reviewed the code also told ProPublica that Lavingia did not clearly define many critical terms, such as 'core medical/benefits,' and used vague instructions, leading to multiple critical contracts being flagged as 'munchable.' For example, the model was told to kill DEI programs, but the prompt failed to define what DEI was, leaving the model to decide. At another point in the code, Lavingia asked the AI to 'consider whether pricing appears reasonable' for maintenance contracts, without defining what 'reasonable' means. In addition, the AI was created on an older, general purpose model not suited for the complicated task, which caused it to hallucinate, or make up, contract amounts, sometimes believing they were worth tens of millions as opposed to thousands. Cary Coglianese, a professor at the University of Pennsylvania who studies governmental use of AI, told ProPublica that understanding which jobs could be done by a VA employee would require 'sophisticated understanding of medical care, of institutional management, of availability of human resources' – all things the AI could not do. Lavingia acknowledged the AI model was flawed, but he assured ProPublica that all 'munchable' contracts were vetted by other people. The VA initially announced, in February, it would cancel 875 contracts. But various veteran affairs advocates sounded the alarm, warning that some of those contracts related to safety inspections at VA medical facilities, direct communications with veterans about benefits, and the VA's ability to recruit doctors. One source familiar with the situation in the department told the Federal News Network that some cuts demonstrated a 'communication breakdown' between DOGE advisors, VA leaders, and lawmakers who oversee the VA. The VA soon walked that number back, instead announcing in March it would cancel approximately 585 'non-mission-critical or duplicative contracts,' re-directing around $900 million back to the agency. Lavingia, who was fired from DOGE approximately 55 days his blog and released the code he used at the VA on GitHub.
Finextra
8 hours ago
- Finextra
Tech-Driven BNPL: How Sophisticated Technologies Are Reshaping the BNPL Market: By Bekhzod Botirov
Bekhzod Botirov, fintech expert, co-owner and member of the PayWay Supervisory Board, outlines how new technologies are reshaping the global BNPL market from reducing risks and improving customer services to refining operations and providing increasingly sophisticated offerings. By 2028, the number of users of BNPL services (Buy Now, Pay Later) is predicted to double to 670 million, an explosive 107% growth compared to 2024. However, as the industry flourishes, so inevitably do the risks ranging from fraud to late payments. To address these issues, international leaders such as Klarna, Afterpay, PayPal, and Affirm are already using artificial intelligence (AI) and big data to minimise their losses and at the same time personalize services for customers and increase sales. Affirm has introduced dynamic payment schedules in the US, while Riverty in Germany uses AI-driven tools to predict user behavior and optimize repayment plans. Afterpay is using big data and AI to ensure a smooth user experience and improved risk management. PayPal's BNPL solution, Pay in 4, incorporates sophisticated fraud prevention technology and machine learning models to assess creditworthiness quickly. Among other things, Sezzle is using machine learning for customer risk assessment and to offer tailored financing options. These, and other BNPL firms are demonstrating how technology, including machine learning, AI and predictive analytics are being used to make services faster, more secure, and more personalized for consumers. However, the wider context is competitive pressures, regulatory demands, and new standards, all of which are pushing providers to improve credit assessment capabilities. As BNPL companies incorporate technologies to meet improved credit assessment objectives they're also discovering further advantages such as improved fraud detection, flexible, transparent payment options, interest-free payment plans, reduced risk of late payments and so on. And with market growth firmly on an upward trajectory BNPL's early adopters are gaining material and market advantage. In a recent report ResearchandMarkets says the BNPL payment market is expected to grow by 13.7% on an annual basis to reach US$560.1 billion in 2025. Further the global market is forecast to grow at a CAGR of 10.2% during 2025-2030 and by the end of 2030 is expected to be worth approximately USD 911.8 billion. There are also further potential technology driven benefits that may not be immediately obvious. For instance, if technology is used to establish information sharing across BNPL players, all companies will be able to see if a borrower has installment plans with other BNPL companies, making the market more transparent and significantly reducing defaults. The growth of BNPL is directly tied to advancements in digital payment technologies, making them an inseparable part of the market's future, so at the very least awareness of the potential of new technologies is incumbent on all players as the market continues to evolve. AI powerfully improves operations from scoring to personalisation AI is having a dramatic impact on the BNPL market. AI-powered credit systems reduce default rates and improve customer satisfaction. Providers that excel in data-driven decision-making will strengthen their market leadership with in-depth analysis of customers' financial behaviour such as what they spend money on, what they invest in, how often they take out loans or request a credit history. Tied to AI are neural networks which can, among other things, also assess a user's social media behaviour to provide ever deeper insight into 'credit worthiness'. AI algorithms can even consider macroeconomic factors like rising unemployment in different regions. AI can also help predict the probability of defaults by detecting patterns that indicate possible financial difficulties such as unstable payments on other instalment plans. It can also help improve customer experience and reduce employee workload and service costs. For instance, AI assistants can carry out the initial processing of customer requests, automate the collection of debts and send borrowers reminders about payments as well as updating customer information. AI can also help personalise offers for users and increase conversions. If, for example, a borrower is making payments on time, customised repayment schedules or raised borrower limits can be offered. It's also possible to predict which product instalment will be the most relevant for the customer. If a consumer bought a PlayStation several years ago, a trade-in programme can offer a new model. For fraud prevention, neural networks can identify anomalies such as a customer applying for a new line of credit from a location that is different to the usual location. Machine learning models can identify high-risk borrowers, fraudulent activities, and outlier behavior. BNPL market leaders are already actively using AI. Klarna and Riverty have implemented machine learning models to offer personalised payment schedules and identify high-risk borrowers. Klarna has also partnered with OpenAI to launch an AI assistant. In its first month alone, it had 2.3 million conversations with customers, two-thirds of all dialogues. The company claims that the bot does the work of seven hundred full-time employees. But AI isn't the preserve of international market leaders. Alif, an Uzbek company, has developed a machine learning based credit scoring model that reduces the time to make decisions on applications to seconds, reduces the percentage of delinquencies and increases the sales of goods in instalments. Alif has also introduced a chatbot that handles thousands of consumer queries across different communication channels, far faster than people could. Blockchain, a new world of transparency and financing models The use of blockchain technology is still in its early stages, but it holds significant potential to transform various aspects of BNPL operations, from improved transparency and trust to regulatory compliance. For instance, it eliminates the manipulation of records of payments, debts, and transaction terms, as each transaction is recorded in a distributed ledger. Blockchain also allows many processes to be automated through smart contracts. These digital agreements are honoured automatically when conditions are met. As an example, if a customer is severely late with a payment, a smart contract can activate sanctions. BNPL platforms can also use smart contracts that automatically analyse a user's wallet and provide a score based on machine learning algorithms. The analysis considers the transaction history in the blockchain such as cryptocurrency payments and activity on DeFi platforms. With the help of blockchain, BNPL services will also be able to raise finance. Tokenised assets backed by receivables can be issued. Investors will buy them on secondary markets, increasing the liquidity of BNPL providers. And cryptocurrencies can facilitate cross-border transfers and help companies receive capital from investors around the world without the complexities of currency regulation. That said, the risks of using unstable cryptocurrencies, such as Bitcoin, needs to be noted. Fluctuations in value can affect the size of the debt. The solution in this case could be stablecoins, the rate of which is linked to other assets. Nexo, a large international company, uses this method to save crypto assets, pay with them and take loans. Nexo claims that the volume of transactions and loans issued on the platform has already exceeded $320 billion. In order to develop the market, government agencies need to develop a legal status for BNPL players on the blockchain. But it's important to note its early days for blockchain. There are not many specialists who know how to develop blockchain systems, even in the global market. For instance, international BNPL services are still looking at the technology. Klarna only announced in February of this year that it was exploring options for integrating cryptocurrencies into its platform. While blockchain offers enormous potential today blockchain adoption is more complicated than AI. A number of regulatory and infrastructural issues need to be resolved to develop the technology. The most realistic scenario today is the development of hybrid BNPL services. In this case, the currency familiar to the population, and blockchain technologies, can be used to record and automate payments. But for this purpose it is still necessary to create a local platform supporting smart contracts for BNPL. Road to the future is lined with superapps and cards In Asia-Pacific, BNPL adoption is heavily influenced by integration with super apps like Grab, Gojek, and WeChat. These platforms offer instalment plans across various services, from ride-hailing to food delivery, providing users with a single app to access myriad services. Superapps serve millions of users daily, so it makes absolute sense for BNPL providers to use these platforms to gain instant access to a vast, engaged audience. It also makes sense for the superapp platform. By embedding BNPL, these apps increase user engagement and transaction volume across multiple services. For instance, Grab PayLater provides BNPL services to millions of Grab users for rides, food delivery, and online shopping. The Paytm Postpaid superapp in India uses Paytm's transaction data to determine BNPL eligibility. And in China, Alipay and WeChat Pay offer BNPL options that allows users to split payments across thousands of merchants. BNPL providers can offer personalized credit limits, reduce default risks with better scoring models and provide custom BNPL plans based on user history. BNPL services integrated into superapps also allow providers to provide instant checkout options, loyalty programs and cashback offers and embedded financing across multiple services. International BNPL leaders such as Klarna, Affirm and Afterpay, in partnership with commercial banks, marketplaces, e-commerce shops and large retail chains, also offer debit cards to users. They can be used, among other things, to buy goods in instalments. However, there is certainly potential to offer even more services such as providing points for on-time instalment payments, which consumers can spend on real goods. Looking further ahead, banks, including microfinance banks, could cooperate with specialised BNPL services and issue debit cards on a white label model. Of course, this approach would require adherence to regulation and would probably require licences from BNPL-providers. Superapps are reshaping the BNPL landscape by embedding BNPL into everyday digital experiences. Their massive user bases and data insights make them 'goldmine' partners for BNPL providers. At the same time cards have a bright future in some territories, and while already in widespread adoption there is certainly room for added services that refine BNPL offerings.
