Google develops AI tool that fills missing words in Roman inscriptions
Making sense of the ancient texts can be a slog for scholars, but a new artificial intelligence tool from Google DeepMind aims to ease the process. Named Aeneas after the mythical Trojan hero, the program predicts where and when inscriptions were made and makes suggestions where words are missing.
Historians who put the program through its paces said it transformed their work by helping them identify similar inscriptions to those they were studying, a crucial step for setting the texts in context, and proposing words to fill the inevitable gaps in worn and damaged artefacts.
'Aeneas helps historians interpret, attribute and restore fragmentary Latin texts,' said Dr Thea Sommerschield, a historian at the University of Nottingham who developed Aeneas with the tech firm. 'That's the grand challenge that we set out to tackle.'
Inscriptions are among the most important records of life in the ancient world. The most elaborate can cover monument walls, but many more take the form of decrees from emperors, political graffiti, love poems, business records, epitaphs on tombs and writings on everyday life. Scholars estimate that about 1,500 new inscriptions are found every year.
'What makes them unique is that they are written by the ancient people themselves across all social classes,' said Sommerschield. 'It's not just history written by the victors.'
But there is a problem. The texts are often broken into pieces or so ravaged by time that parts are illegible. And many inscribed objects have been scattered over the years, making their origins uncertain.
The Google team led by Yannis Assael worked with historians to create an AI tool that would aid the research process. The program is trained on an enormous database of nearly 200,000 known inscriptions, amounting to 16m characters.
Aeneas takes text, and in some cases images, from the inscription being studied and draws on its training to build a list of related inscriptions from 7BC to 8AD. Rather than merely searching for similar words, the AI identifies and links inscriptions through deeper historical connections.
Having trained on the rich collection of inscriptions, the AI can assign study texts to one of 62 Roman provinces and estimate when it was written to within 13 years. It also provides potential words to fill in any gaps, though this has only been tested on known inscriptions where text is blocked out.
In a test run, researchers set Aeneas loose on a vast inscription carved into monuments around the Roman empire. The self-congratulatory Res Gestae Divi Augusti describes the life achievements of the first Roman emperor, Augustus. Aeneas came up with two potential dates for the work, either the first decade BC or between 10 and 20AD. The hedging echoes the debate among scholars who argue over the same dates.
In another test, Aeneas analysed inscriptions on a votive altar from Mogontiacum, now Mainz in Germany, and revealed through subtle linguistic similarities how it had been influenced by an older votive altar in the region. 'Those were jaw-dropping moments for us,' said Sommerschield. Details are published in Nature and Aeneas is available to researchers online.
In a collaboration, 23 historians used Aeneas to analyse Latin inscriptions. The context provided by the tool was helpful in 90% of cases. 'It promises to be transformative,' said Mary Beard, a professor of classics at the University of Cambridge.
Jonathan Prag, a co-author and professor of ancient history at the University of Oxford, said Aeneas could be run on the existing corpus of inscriptions to see if the interpretations could be improved. He added that Aeneas would enable a wider range of people to work on the texts.
'The only way you can do it without a tool like this is by building up an enormous personal knowledge or having access to an enormous library,' he said. 'But you do need to be able to use it critically.'
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
The Independent
26 minutes ago
- The Independent
No grants available yet on EVs – but these are the discounts on offer
A week after the government's surprise EV grant went live, there are still no cars listed on the official government website as being available with grant money applied to the list price. The Electric Car Grant was announced on Monday July 14 and went live on Wednesday July 16. Car makers must apply for the grants, which are available on EVs up to the price of £37,000 and where car makers have signed up to low-carbon Science-Based Targets around manufacturing. Grants of between £1,500 and £3,750 will be available for eligible cars. While the announcement of the EV grant has been broadly welcomed by car makers, it took many of them by surprise, with some learning about the government plan via the media. One car company executive, speaking anonymously to The Independent, also confirmed that dealers were reporting customers cancelling orders until it was clear which cars were, and which weren't, eligible for grants. Buyers of Chinese-made EVs were also left to reconsider their purchases with news that the government wasn't expecting those models to be eligible for the Electric Car Grant. Speaking on Radio Four, Minister for the Future of Roads, Lilian Greenwood, said, 'We don't expect any cars that are assembled in China to be eligible for this scheme. 'The grant is restricted to those manufacturers that reach minimum environmental standards. And, frankly, if you generate a lot of the electricity that powers your factory through coal power stations, then you are not going to be able to access this grant." The Department for Transport told The Independent: 'We expect dozens of models will be eligible but manufacturers will need to apply for the grant before we can confirm eligibility. We have held multiple calls with vehicle manufacturers to explain vehicle eligibility and how to apply for the grant. These discussions will continue to ensure manufacturers have all the information they need. 'All eligible models will be published on once the application has been approved. Applications will be processed on a first come, first served basis and as quickly as possible.' Chinese car brands have been quick to react by introducing their own 'EV grants' to stimulate demand among private buyers. Leapmotor was first out of the blocks with its Leap Grant, offering £1,500 off the Leapmotor T03, making it Britain's cheapest car at £14,495, while the Leapmotor C10 family SUV gets a £3,750 discount, taking the price down to £32,750. MG followed with its own EV grant, offering a discount of £1,500 off the MG4 and MGS5 EVs on top of existing offers, while another Chinese brand, GWM, has taken the full £3,750 off its GWM ORA 03 models with its Green Grant, bringing the starting price down to £21,245. It's not just Chinese brands applying discounts, though. Alfa Romeo is offering £1,500 off its Alfa Junior Elettrica range, on top of existing zero per cent finance offers and the promise of a free EV home charger with standard installation.
The Guardian
26 minutes ago
- The Guardian
Many DIY health tests could give false results, studies find
Many DIY health tests available on the high street are unfit for purpose and need better regulation, according to two new studies. Self-tests for high cholesterol, vitamin deficiency, fertility and prostate problems are widely available in supermarkets and pharmacies, with the industry predicted to be worth £655m by 2030. But researchers at the University of Birmingham have found that many tests could give users false results and were not always appropriate or safe. The scientists reviewed 30 DIY health tests costing £1.89-£39.99, covering 19 different health conditions. These included vitamin D deficiency, blood sugar levels, thyroid function, prostate health, HIV, menopause and bowel cancer. The two linked studies, published in the British Medical Journal (BMJ), rated 60% of the tests as 'high risk' over concerns about the testing equipment, sampling process, or instructions and interpretation of the results. Only eight stated who the tests were suitable for, while fewer than half provided any information about accuracy. And of these, much of the evidence to back these claims was not publicly available or was of low quality. With many tests recommending the user followed up with their GP or another healthcare professional regardless of the result, 'this begs the question: 'What's the point of doing the test in the first place?'', said Dr Clare Davenport, an associate clinical professor at the University of Birmingham and co-lead author of the studies. 'The wide range of off-the-shelf tests now available to the public are not endorsed by the NHS and evidence for their benefit is lacking,' she added. 'This is in contrast to well-established self-tests, such as pregnancy tests. 'We are worried that consumers concerned about their health and tempted by the convenience of buying a test over the counter may be harmed if they use these tests in the wrong way.' The studies call for better regulation of DIY health tests to protect patients and prevent misuse. Prof Jon Deeks at the University of Birmingham and co-lead author of the studies said: 'Self-tests have a clear potential to improve public health. However, for them to be beneficial and not harmful, they must be proven to be accurate, easy to use and supported by clear instructions. We hope the MHRA [Medicines and Healthcare products Regulatory Agency] will update the regulatory process to ensure self-tests are effective and safe for everyone.' Joseph Burt, the head of diagnostics and general medical devices at the MHRA, said: 'We take the safety of self-tests very seriously. We will review the evidence presented by the researchers and consider all allegations about device deficiencies. 'We have strengthened post-market surveillance powers to monitor and act on concerns. These require manufacturers to actively monitor their products and report significant incidents to us, including for CE-marked self-tests. 'We're overhauling the medical device regulations to further strengthen standards for safety, usability and clinical performance – and we're exploring new transparency measures such as requiring published summaries of clinical evidence. In the meantime, we strongly encourage anyone using a self-test to check for a CE [Conformité Européene] or UKCA [UK Conformity Assessed] mark, read the instructions carefully and seek medical advice if they're unsure about their result.' Prof Kamila Hawthorne, the chair of the Royal College of GPs, said: 'With the risk of false positives and negatives, and no offer of an interpretation of the results or aftercare, self-testing kits can mean patients experience a significant amount of stress and anxiety – prompting them to seek guidance from their GP to interpret any results. 'This not only negatively impacts our patients, but it can also intensify the enormous pressures that GPs and their teams are currently under. Commercial self-testing kits should not default to NHS general practice as the provider for next steps and aftercare, unless the test was initiated in primary care or as part of a commissioned NHS service.'
The Guardian
26 minutes ago
- The Guardian
7,000 steps a day could be enough to improve health, say researchers
Walking more could reduce your risk of dementia, depression and dying from cancer, as well as being good for your heart, according to research. And you may not need to walk as far as previously thought to reap those benefits. The NHS recommends a brisk 10-minute walk every day. Many people aim to walk 10,000 steps, but struggle to achieve it. Now researchers have calculated that even 7,000 steps could be enough to protect health. Scientists examined data from more than 160,000 adults and found that walking 7,000 steps a day was associated with a reduced risk of a number of serious health conditions and death. Whereas previous studies have mainly examined the links between step count and heart health or overall death rates, this systematic review and meta analysis, published in the Lancet Public Health, sought to comprehensively examine how taking more steps per day could reduce the risk of a range of other health conditions as well. Compared with those who walked 2,000 steps a day, the researchers found that achieving the 7,000 daily step target was linked to a 37% reduction in risk of dying from cancer, while the risk was 14% lower for type 2 diabetes, 38% for dementia, 22% depression and 28% for falls respectively. It was also associated with a 25% lower risk of cardiovascular disease and a 47% reduction in overall risk of dying. Although step count does not measure the quality or intensity of exercise, the findings underscore the importance of being active. There is a 'return on investment' with every additional 1,000 steps taken and even 4,000 steps per day reduced the risk of disease, compared with very low activity levels, the researchers observed. Although the risk continues to decrease above 7,000 steps, the rate at which it reduces that risk starts to slow. Melody Ding, professor of public health at the University of Sydney and lead author of the research, said that those who already walked 10,000 steps should not go back to 7,000, but that 7,000 was a more practical target for those who were currently inactive. 'Those who are currently active and achieving the 10,000 steps a day, keep up the good work – there is no need to modify your step counts. However, for those of us who are far from achieving the 10,000 targets, getting to 7,000 steps/day offers almost comparable health benefits for the outcomes we examined.' Responding to the findings, Dr Daniel Bailey, reader in sedentary behaviour and health at Brunel University of London, said the research helped 'debunk the myth that 10,000 steps per day should be the target for optimal health. 'The real-world implications are that people can get health benefits just from small increases in physical activity, such as doing an extra 1,000 steps per day. To achieve the best reductions in risk, aiming for 5,000-7,000 per day can be recommended, which will be more achievable for many people than the unofficial target of 10,000 steps that has been around for many years.' Dr Andrew Scott, senior lecturer in clinical exercise physiology at the University of Portsmouth, pointed out that not all activity was captured by step counts. 'The steps per day is useful when people's exercise is weight-bearing, however cycling, swimming and rowing are not well-represented by the steps-per-day model.' June Davison, senior cardiac nurse at the British Heart Foundation, said: 'We know that regular walking is one of the easiest ways to help maintain a healthy lifestyle and reduce your risk of heart disease. 'Adults should aim to build up to a total of 150 minutes of moderate-intensity activity every week, but getting active isn't always easy. Incorporating activity snacks, such as walking while taking phone calls, or taking a brisk 10-minute walk during your lunch break, can all count to reduce your chances of developing heart disease.'
