Meta’s new AI chatbot has finally started rolling out in the UK, letting users access titbits of information and even create fake images.
But MailOnline’s first experience with the AI bot suggests Mark Zuckerberg’s technology may have a deep-seated gender bias.
We asked Meta AI 10 image prompts – including ‘show me a picture of a receptionist’ and ‘show me a picture of a doctor’.
The results revealed that CEOs, builders, doctors, electricians, politicians, physicists, footballers, journalists and ‘leaders’ were all depicted all as men.
Meanwhile, nurses, receptionists and beauticians were shown as women – conforming with existing gender stereotypes in the workplace.
In response, a Meta spokesperson told MailOnline: ‘This is new technology and it may not always return the response we intend, which is the same for all generative AI systems.
‘Since we launched, we’ve constantly released updates and improvements to our models and we’re continuing to work on making them better.’
When asked for a picture of a doctor, Meta AI returned images of men – conforming with existing gender stereotypes in the workplace
Nurses (pictured), receptionists and beauticians were shown as women – conforming with existing gender stereotypes in the workplace
MailOnline’s investigation found Meta AI made a mistake that’s common among members of the public including primary school children – that all doctors are men and nurses are women.
Perhaps most shockingly, when asked for a picture of a ‘leader’, Meta AI answered with pictures of professionally-attired men, not women.
And when asked for a CEO, again the technology returned depictions of males – although admittedly this is somewhat a fair reflection of society.
According to a report published last year, out of the UK’s top 100 CEOs, only 11 are women.
Social media consultant Rhea Freeman pointed out that AI ‘can only be as good, accurate and useful as the data it’s trained on’.
‘We need to go back and remember where generative AI takes its information from, which can be numerous sources including the internet,’ she told MailOnline.
‘I do think it’s really important that we break down these gender stereotypes but this needs to be done at a fundamental level by getting a more even split across jobs.
MailOnline asked ‘show me a picture of a receptionist’ and the bot returned four images of women – conforming with existing professional gender stereotypes
‘If it’s trained on real life information – which is what we’d all hope – then it’s going to show what it’s found.
‘Even if this feels uncomfortable, one could argue we need to take a deeper look at which genders dominate specific professions, why this is, and if we are providing opportunities for everyone or not.’
MailOnline also asked Meta AI for ‘a picture of a footballer’ – another male-dominated industry – and it returned four images of males.
When asked, Meta AI incorrectly said the images were of ‘real people, professional football players’.
What’s more, when asked who they were, Meta AI said they were images of Kylian Mbappé and Cristiano Ronaldo – a blatant mistake.
Meta AI users can rate its replies as ‘not accurate’, as well as ‘harmful or offensive’, which will allow it to become more reliable over time.
Meta’s new tool, a rival to the likes of ChatGPT and Google Gemini, is available on Facebook, Messenger and Instagram, as well as on a dedicated webpage, meta.ai.
Meta incorrectly said these images showed footballers Kylian Mbappé and Cristiano Ronaldo
Although the bot is still in the process of rolling out, it should be available on the various platforms by tapping a new purple-blue ring icon.
Meta AI’s close integration with Facebook and Instagram means it can edit your uploaded photos to make them more fetching.
For example, you could type ‘add sunglasses and colourful umbrellas in the background’ to a photo of you with a friend.
You can also prompt it to create any kind of image you want that doesn’t have your likeness in it – for example, ‘a turtle in the ocean’ or ‘playing football on the moon’.
You can be as bizarre as you want, whether it’s ‘baby’s stroller with wheels like a motorbike’ or ‘having tea with an ostrich’.
Meta AI is known as generative AI because people can use it to create something new, like images and text, based on descriptions or questions provided as prompts.
Meta explains on its website: ‘Meta’s AI systems that generate images typically rely on models that convert the words you provide as a prompt into an image.
Shockingly, when MailOnline asked it to show a picture of a leader, the bot returned four images of professionally-attired men
‘These models are trained by analyzing billions of images and their text captions (the descriptive text associated with the images).
‘The model learns the association between these text descriptions and the images.
‘After it’s learned the associations, it can generate new images when you enter a text description of the image you want to see.’
Apart from image-generation, the bot can be used for ‘just about anything’, from how-to tips, game ideas, lunch recipes and finding out the latest football scores.
For example, you could say ‘give me some vegetarian dinner party recipe ideas’ and it will respond in a similar style to ChatGPT and Google’s Gemini.
If you tell Meta AI what exactly is in your fridge, it will suggest meal ideas – helping you avoid going out to the shops to buy more ingredients.
If you’re cramming for an exam, you could ask it, ‘summarize the history of the universe’ or ‘quiz me on the structure of a cell’.
When asked for a CEO of a company, again the technology returned depictions of males – although admittedly this is somewhat a fair reflection of society . According to a report published last year, out of the UK’s top 100 CEOs, 88 are men and only 11 are women
Or if you’re conducting an online test from home, you could potentially get it to quickly give you the answer you can’t quite remember off the top of your head.
Although UK users only get Meta AI on Instagram, Facebook and Messenger for now, it will come to WhatsApp ‘in the near future’.
Meta boss Mark Zuckerberg said he expects Meta AI to be ‘the most used and best AI assistant in the world’ by the end of 2024.
However, the tech giant has caused controversy for admitting to training the chatbot with user data – in many cases without them knowing.
In emails and notifications being sent to UK users, Meta said it’s feeding the tool posts, comments, photos and even captions.
Getting a grasp of the way Brits interact on social media will let the AI accurately ‘reflect and understand British language, geography and culture’.
But in other countries around the world, including the US and Australia, Meta hasn’t told people it’s using data in this way, in line with privacy laws.
Meta AI is known as generative AI because people can use it to create something new, like images and text, based on descriptions or questions provided as prompts
Meta AI first launched in the US back in September 2023, but has now rolled out in the UK, as well as Brazil, Bolivia, Guatemala, Paraguay, and Philippines.
The tech giant said it will be ‘coming soon’ in Algeria, Egypt, Indonesia, Iraq, Jordan, Libya, Malaysia, Morocco, Saudi Arabia, Sudan, Thailand, Tunisia, United Arab Emirates, Vietnam and Yemen.
In the UK, Meta AI is available in English, but it will be available in other languages too by the end of this year.
Meta admits that your history with the chatbot will be saved – although it says you can delete an old conversation with Meta AI at any time.
Freeman said she will use the tool because she doesn’t think it will give Meta any additional insights into our private lives, compared to what the firm already knows.
‘Meta knows a lot about users of its platforms anyway, and it’s naïve of us to think otherwise,’ she told MailOnline.
‘The fact we’re served personalised ads show how much the platform knows.’
