Google starts rolling out Bard, its answer to ChatGPT
Google is the latest to join the generative AI fray
Tech companies are falling over each other to push AI products, but are safety and ethics the first casualties?
Google has started rolling out Bard, its answer to ChatGPT, to a limited number of users in the US and the UK.
In a blog post on Tuesday the company invited people to sign up to test out its generative AI chatbot, which for the moment is only available in the English language and to over 18s, in order to provide feedback to aid its development.
Perhaps learning from the mistakes made by OpenAI's ChatGPT, which became famous almost as much for its errors as its groundbreaking capabilities, and also Bard's less than stellar initial public outing, Google calls its chatbot an "early experiment" and admits that it will get things confidently wrong - a process known as hallucinating.
The post's authors Sissie Hsaio and Eli Collins provide an example where Bard suggests "Zamioculcas zamioculcas" as a houseplant that's easy to care for, whereas the scientific name for the ZZ plant is actually Zamioculcas zamiifolia. Zamioculcas zamioculcas is an invention.
"While LLMs [large language models] are an exciting technology, they're not without their faults," the authors note. "For instance, because they learn from a wide range of information that reflects real-world biases and stereotypes, those sometimes show up in their outputs. And they can provide inaccurate, misleading or false information while presenting it confidently."
Bard is based on a LLM called LaMDA, which Google has been developing since 2020. Perhaps wary of cannibalising its most profitable products, the company says it will not be added to Google Search or used in advertising.
Like ChatGPT, Bard is not updated live via the web so it has no knowledge of recent events. For now, it is text-only, and unlike ChatGPT will not write code, although this is in the pipeline.
"We'll continue to improve Bard and add capabilities, including coding, more languages and multimodal experiences," the blog post says.
AIs are multiplying
Bard joins a growing list of consumer AI services based on LLMs. Tech companies are rushing to outdo each other in a field where a four-month head start represents a significant margin.
Last week, OpenAI released a premium version of ChatGPT based on its upgraded LLM GPT-4 which can understand images and also process eight times more words than its predecessor GPT-3 and also makes fewer errors, the company claims. Microsoft, which has a multibillion dollar stake in OpenAI, has integrated this version into the Bing search engine.
Not to be left out, in February Meta unveiled its latest LLM LLamA, which it said it planned to open source, only to have it leaked online. Meanwhile, Baidu has released a Chinese language AI chatbot called Ernie.
The rush to release new AI products has raised concerns, particularly as the companies involved have been laying off AI ethics and safety teams in repeated rounds of job cuts. AI models are only as good as the data they're trained on, and the biases inherent in that data are well known. OpenAI's CEO Sam Altman, even professed himself recently to be "a little bit scared" about its potential use for spreading large-scale disinformation and enabling cyberattacks.
In a blog post on Tuesday, Microsoft co-founder Bill Gates described AI as a development as important as the internet with much potential for good, but he said governments and regulators will need to step in to prevent AI from widening existing divisions.
"Market forces won't naturally produce AI products and service that help the poorest. The opposite is more likely," Gates wrote.
"With reliable funding and the right policies, governments and philanthropy can ensure that AIs are used to reduce inequity. Just as the world needs its brightest people focused on its biggest problems, we will need to focus the world's best AIs on its biggest problems."