OpenAI releases ‘not fully reliable’ tool to detect AI generated content
- OpenAI have released a new classifier tool designed to detect whether text has been written by AI.
- The tool has been trained to distinguish between text that has been written by AI and text written by a human.
- They have warned that it is not completely reliable yet and whilst it is “impossible to reliably detect all AI-written text”, good classifier tools could pick up signs that the text was written by an AI system.
- OpenAI states that the reliability of the tool improves as the length of the input text increases.
- The main limitations, as reported by OpenAI, include its unreliability on text below 1000 characters, as well as misidentification of some human-written text as AI-written.
- For more on this, please visit the Guardian’s website.
Call to stop rogue algorithms ‘pushing’ antisemitism online
- The Antisemitism Policy Trust has highlighted how ‘unfit for purpose’ systems can spread disinformation and harmful content through search engines.
- The Trust’s chief executive welcomed the Online Safety Bill as a first step to this. He said, “I’ve always been of the view that the bill should address the systems behind social media and internet platforms, rather than the content.”
- He suggests that home voice technologies such as Siri, Hey Google and Alexa, should also fall under the regulation.
- Furthermore, he urges that protection from harmful content should be the default setting rather than an opt-in feature and that people shouldn’t have to see content on antisemitism and holocaust denial, unless they actively seek to.
- For more on this story, please visit the Independent’s website.
Learning loss for disadvantaged pupils ‘greater’ than peers, says watchdog
- The National Audit Office (NAO) have warned that if unaddressed, lost learning could lead to “increased disadvantage and significant lost earnings” for those affected.
- The NAO have stated that disadvantaged pupils in England “remain further behind” the expected level of achievement than their peers.
- Research commissioned for the Department for Education (DfE) in Summer 2021 found that disadvantaged secondary school pupils were around 2.4 months behind where they were expected to be in reading, compared with 1.2 months for all other secondary pupils.
- The report also shows that around half of the pupils receiving tutoring under the Government’s National Tutoring Programme (NTP) were disadvantaged in 2021/2022.
- The head of the NAO, Gareth Davies, has stated “The Department for Education needed to take action to support pupils to make up the learning they lost during the Covid-19 pandemic and reach children who had been disproportionately affected by the disruption to schooling.”
- The DfE have stated they are investing more than £5 billion over the next two years to improve outcomes for disadvantaged children across England.
- For more on this story, please visit ITV News website.