AI and Human Expectation: A Treatise

Artificial intelligence (AI) is drawing attention once again. Technologies such as DALL-E and Stable Diffusion are letting users generate images from text prompts, and ChatGPT is the latest in a long line of popular chat bots. The reception of these services has bolstered the fortunes of OpenAI, which is reported to have received a $10 billion investment from Microsoft. The tech giant previously gave OpenAI $1 billion in 2019, as part of a deal to help build Microsoft Azure’s AI supercomputing platform. OpenAI would then exclusively use Azure to run its services.

Despite the sense of novelty about these products, AI and machine learning tech is already widespread. Virtual assistants such as Siri and Alexa live on the cloud as well as in smartphones and smart speakers. Google Translate on Android offers on-device translation for over 50 languages without needing a network connection — especially helpful for travellers who don’t always have access to mobile data.

These services are useful for people looking for an answer to a problem. Really, that’s the basic definition of any technology: a tool for solving a problem.

But DALL-E and ChatGPT are also thought to be useful for content creation. DALL-E could be a comparatively inexpensive alternative to stock photography, and ChatGPT has been used by students to write essays — a new frontier in academic dishonesty. Some news outlets are even using AI to write articles, as it allows for the quick development of highly targeted content to improve search rankings, as well as the ability to be the first with a story. AI also works out cheaper than paying human writers. All this raises serious ethical questions.

Slave to the Algorithm

Using AI to write news articles is fundamentally an exercise in search engine optimization (SEO), the decades-old practice of maximizing keyword and link placement to get a higher spot in Google’s results. For an ad-supported sales model this can be make-or-break. It’s also lucrative; in 2020 Google sent people to news sites over 24 billion times a month. Although there’s an extent to which SEO can be used for opinion pieces and current events, for evergreen topics like “what is a credit score?” or “how do I unclog a drain?” it’s crucial for good placement.

The above topics are also low-hanging fruit for news websites using AI to write. The premise looks attractive from a management perspective — an endless stream of content relevant to your audience, bundled with ads to generate revenue, that users can scroll continuously through. In theory it makes sense to use AI in these scenarios, because it solves the “problem” of paying writers.

But what if that first click never occurs? If the technology to answer questions of varying complexity exists, and people use search engines to find answers, why would a search engine direct someone to an external location when it can provide an answer itself?

The precedent is already set. Solving maths problems in search queries was supported from the earliest days of Google. Other calculations, such as unit conversion, followed soon after. Information panels exist for various people, places, organizations and things, and searches for lyrics return the complete songs. Applying AI to search engines is a natural progression of both technologies.

Searching for a New Old Business Model

Consider that many searches, particularly from non-technical users, are simply asking a question. Some of these questions refer to a matter of fact, like “where do I register to vote?”. AI could answer this definitively with a single, potentially personalized, answer. Other questions refer to a matter of taste, for example “how do I make spaghetti carbonara?”, which an AI could answer subjectively.

Using natural language in search queries was the initial conceit of Ask Jeeves, one of many search engines from the late 1990s that struggled to compete with Google. An updated take on the concept, using modern AI, may pose much more of a threat.

Microsoft’s investment in OpenAI could present an opportunity for its highly promoted, yet scarcely used, Bing search engine to gain relevance. Bing could use AI to give rich results that answer users’ questions without directing them to an external site. Microsoft is reportedly already working on integrating OpenAI’s technology with its Office platform.

At CES 2023 earlier this month, Microsoft’s chief product officer Panos Panay exclaimed that “AI is going to reinvent how you do everything on Windows”. Subscribers can read more about the event here; for more information please get in touch. As Microsoft and OpenAI have strengthened Azure’s AI capabilities together, the prospect of OpenAI coming to Bing isn’t a stretch — but it could be threat to Google’s search business.

On the content side, the prospect of creating AI-generated content libraries to aid SEO may provide short-term benefits, but it’s unlikely to be useful in the long run. The introduction of AI-answered questions wouldn’t eliminate links in searches, but it would push search results further down the page.

Despite valid arguments about the accuracy of ChatGPT and the value added by human editors, the tech is fundamentally still a prototype, although given the established speed of release of OpenAI, it’s likely to improve very soon.

What makes news and commentary ultimately stand out remains the same: boots-on-the-ground reporting, writers with strong opinions and a voice that draws engagement.