Prompted: It’s Time for a Better AI Model

If Isaac Newton were to write a law for AI news, it would go something like this: With any new AI release, there will be an equal amount of backlash in the opposite direction. From concerns over energy consumption to plagiarism accusations, tech companies are finally taking these issues into account and shifting how they produce AI models. The move towards smaller, more focused, and less copyright-infringing AI models marks a noticeable change in the AI landscape — and could also affect how you use them.

Smaller, Sleeker, More Productive

Recently, top tech companies from OpenAI to Google have been focusing on a new type of AI: the mini model. These compact versions are trained on much smaller datasets and are typically focused on single areas of use, rather than attempting to accomplish every task and answer every question. For example, these models can be trained on a company’s communications to help write emails or on legal documents to assist in case research. Financial services company Experian shifted to smaller models for its customer chatbots, which give financial advice and assist with customer service, and trained them only on internal company information. In an interview with the Wall Street Journal, Experian’s chief data officer Ali Khan said the models “train on a well-defined problem area and set of tasks, as opposed to giving me a recipe for flan.”

Build Your Own Mini Model

If your brand is in the market for its own AI tool, whether for internal use like crafting on-brand marketing messaging or for customer-facing tasks like an advice-giving chatbot, these smaller models might be right for you. The top benefit? Price. Smaller models are much more cost-effective, costing less to develop and saving you money on each prompt, since they require less computing power to run.

But even if going the DIY route is not in your marketing plan (or budget) for this year, these new models can still benefit the everyday user. OpenAI’s GPT-4.o mini can return answers faster than the full version and uses less energy to produce an output (which is good, since AI models are currently spiking power demand and increasingly straining our already overburdened grid).

Plagiarism-Free Content

Ever since the initial AI explosion in 2022, AI users have been consistently concerned about the technology’s shady relationship with copyright and plagiarism. In response, AI companies are now creating models that are a little less lawsuit-prone. 

For OpenAI’s newest search tool, SearchGPT, the AI giant is teaming up with several news organizations to safely incorporate their content into the model. In the recently released prototype, SearchGPT features search results that prominently cite and link to relevant articles, putting sourcing information front and center.  Another popular chatbot, Perplexity, plans to share ad revenue with publishing partners after facing numerous plagiarism accusations from prominent media outlets. AI image generators are also attempting to curb copyright complaints — companies like Shutterstock, Getty, and Adobe are now paying photographers for images used in AI training.

SEO Evolution

AI is already transforming the world of SEO, and these new AI-publisher partnerships only complicate the landscape further. While it’s too early to tell whether this will restore waning traffic, these updates are still a boost for a languishing media industry. 

As PR pros, one thing we’ll be looking for is how these shifts affect the value of brand inclusion in the media. If these search tools take over, it could make the importance of a good media hit all the more valuable. With news articles directly integrated into search, landing a placement in a well-performing pub could give you a better chance of also appearing in search results on topics important to your brand.

A New Path for AI

Overall, the move to smaller, less energy intensive, and less plagiarism-prone models is a positive development for AI skeptics and AI lovers alike. For the hesitant, this could be a step in the right direction, demonstrating that AI companies are taking important consumer concerns into account. For current AI users, these changes could improve optics. Marketers can leverage these models to assure their audience of the safety and reliability of the AI tools being used, enhancing customer trust and showing a commitment to ethical AI standards. And for marketers looking to use AI to boost content creation, this shift can help maintain the integrity of your brand’s voice and avoid potential legal issues.

No matter the use case, this new direction for AI model building shows that tech companies aren’t blind to the backlash they’re receiving. We’ll be keeping our eyes on the industry as additional adjustments are made and new guardrails are put in place for the benefit of all AI users.

Prompted, our biweekly blog series, brings you relevant updates and insights on how to strategically add AI-driven solutions into your marketing toolkit. Subscribe to receive updates in your inbox.