BIG TECH BUYS IN
Chatbot AI seems the game of the name right now. Currently, with over 100 million users, OpenAI’s chatbot, ChatGPT, has set the record for the fastest-growing consumer application in history. Microsoft has invested $10 billion in the technology, planning to integrate it into its ecosystem of products, including Bing and Office 365. Hot on the heels of the Windows vendor, Google launched Bard, a chatbot based on its Language Model for Dialogue Application (LaMDA).
AI THAT DOESN’T MAKE THE GRADE
Despite the hype, both ChatGPT and Bard have user shortcomings. A UK university student aiming for a first on an essay used ChatGPT to do the work. The result was a disappointing 2:2 grade. OpenAI has also admitted that its brainchild can produce biased answers and mixes fact and fiction. Alphabet, Google’s parent lost $100 billion in market value on 8 February after Bard shared inaccurate information in a promo video. So obviously these chatbots aren’t as dazzling as may seem. But mistakes aren’t the biggest problem. With ChatGPT and Bard, Microsoft and Google potentially risk grave peril of legal liability for every bit of misinformation and libel that their inventions create.
DID OPENAI ASK?
Many are asking the question – is it too early to talk about AI and the law? And the answer from just as many is – it’s too late! Without getting into the nitty-gritty of the machine-learning algorithms underlying ChatGPT techniques, realise that all the content that these generative AI systems use was created by humans and scrapped from the web. Almost in the blink of an eye, the chatbot accesses 300 billion words systematically harvested from the internet including books, articles, websites and posts. ChatGPT also hones in on your personal information obtained without consent. This could be a blog post, a product review or a comment on an online article. Were we asked by OpenAI if it could use our data? The straight-up is no. Some say this is a blatant violation of privacy, especially for sensitive data used to identify us, our family members, or our location.
WHOSE DATA IS IT?
Looking at ChatGPT from a different legal perspective, OpenAI did not pay for the data it scraped from the internet. Nobody – individuals, website owners, companies or the government agencies that produced the content have been compensated. Does that seem fair considering OpenAI’s $29 billion valuation?
TEXTUAL INTEGRITY AND COMPLIANCE
When we put our data out publicly on the web, it doesn’t mean that anybody can use it for anything. Underpinning the legal aspects of data privacy is the principle of textual integrity. This means that information cannot be used outside of the context in which it was originally produced or intended without permission. Under the GDPR and other data privacy regulations, everybody retains the “right to be forgotten” particularly important in cases where information is inaccurate or misleading, a regular occurrence with ChatGPT. In this regard, OpenAI lacks transparency and there is no way for us to check how our data is stored or if we can request for it to be deleted. When it comes to data privacy and compliance, ChatGPT is skating on thin ice. Maybe ChatGPT and Bard will be like Facebook where you electronically sign away your right to be forgotten and your personal freedoms without even knowing it.
ONLY FOR HUMANS
Imagine that ChatGPT has created a piece of content that you love. So much so that you want to copyright it. You can forget it. At the moment, for work to enjoy copyright in the United States or the UK it must be created by a human. Even if the work could be copyrighted, what happens when ChatGPT generates duplicate or similar content for another user? Can the initial user sue others for copyright infringement? Margaret Esquenet, a partner at IP law firm Finnegan has this to say:
“Even assuming that the rightful copyright owner is the person whose queries generated the AI work, the concept of independent creation may preclude two parties whose queries generated the same work from being able to enforce rights against each other. Specifically, a successful copyright infringement claim requires proof of copying—independent creation is a complete defence. Under this hypothetical, neither party copied the other’s work, so no infringement claim is likely to succeed.”
WHO IS TO BLAME?
Given the nature of the beast, ChatGPT is capable of generating biased, harmful and damaging content. Who is to blame or legally responsible? IP lawyer, Michael Kelber poses this question
“An exact copy of protected work could create potential liability, which raises another question: who is liable, the creator of the AI — such as ChatGPT — or the user who posed the query?”
AI BLACK BOX
ChatGPT, Bard and other forms of generative AI are fraught with concerns around data privacy, copyright and IP, compliance and even plagiarism. This situation is exacerbated by the fact that these chatbots operate on an input-output model. A user inputs a query and ChatGPT spews out a response. Even the experts at Microsoft and Google will be unable to predict the output. Put simply, nobody knows how these chatbots construct their outputs since neural networks are the quintessential black boxes. Take this extreme example – a user asks ChatGPT for a remedy to an illness. The chatbot provides inaccurate information and the user gets sick. Again, who is to blame? An army of lawyers will provide the answer in seconds. Being able to identify a problem only after the fact won’t be the best defence for Big Tech planning to implement public-facing generative AI.
TEST IT WELL USE IT WELL
While I’m not against using ChatGPT or the like, I am pointing out their immediate limitations. With more than 20 years of experience in the field of professional business IT, I know that everything should be thoroughly tried and tested before becoming a public-facing commodity. If you have doubts about the dos and don’ts of generative AI, please contact me. Tother, we can make the world a safer place online – for everybody.