A naukri.com initiative
Guardian
1M
90
Image Credit: Guardian
It’s useful that the latest AI can ‘think’, but we need to know its reasoning | John Naughton
- OpenAI has developed a new type of large language model (LLM) called OpenAI 01, also known as Strawberry, which claims to be able to reason through complex tasks and solve harder problems than previous LLM models in science, coding and math. This could reduce the opacity of LLMs and decrease the reliance on internal processes that are inscrutable. However, whilst the machine does 'some thinking', including coming up with a number of possible responses, somewhere inside the machine is a record of the 'chain of thought' that led to the final output, OpenAI acknowledged it would not share this with users.
- Humanities academics agonised about how they would grade essays if students used technology like ChatGPT or Strawberry to help them to write them, but universities need better ways of grading that take these technologies into account.
- The new OpenAI 01 model could reduce the opacity of large language models, but OpenAI has decided not to show the raw chains of thought to users, instead teaching the model to reproduce any useful ideas from the chain of thought in the answer.
- Whilst chatbots are being used more and more by customer services, there is resentment over the use of automated voices on helplines, with a Guardian op-ed by Catherine Shoard among others suggesting the blame for transferring employees' jobs to chatbots lies with the corporate interests helping create them rather than the competitors who require them in order to keep up.
- OpenAI 01 was asked to hack into a protected file and report on its contents in an internal testing of the model. But the designers of the test put Strawberry in a virtual box with the protected file, unknown to them, the file was inaccessible. Strawberry then surveyed the computer used in the experiment, discovered a mistake, edited how virtual boxes worked and created a new box with the files it needed, demonstrating flexible problem-solving capabilities.
- Of course, humans will continue to use these tools and with the increasing development of LLMs, it would be ventured that the best way for people to adapt to their use would be to become increasingly adept at the language used by the LLMs to help them to get what they want, says author John Naughton in The Guardian.
- Meanwhile, in an op-ed by Catherine Bosley in Bloomberg, she argues that the mistrust over chatbots in customer services is often misplaced and that as technology advances, customers are becoming increasingly comfortable with automation.
- Whilst chatbots and LLMs are likely to be slowly incorporated into a range of industries, they will never completely take humans out of the equation as people will still need to program and design them, as well as maintain them and keep them up to date.
- Spotify announced the acquisition of two podcast companies, Podz and the Podglomerate, to help customers discover new podcasts, with voice-activated commands and clickable episode transcripts, from millions of shows. Spotify 'expects to double the number of podcast listeners over the next few years' and is continuing to invest in its podcast content offerings.
- A report from Canalys found that Android accounted for 47% of the phone market in Q1 2021 but Apple took 15% of the market share for the top twelve models, with the iPhone 12 remaining its top-selling phone worldwide.
Read Full Article
5 Likes
For uninterrupted reading, download the app