Generative AI v discriminative AI: Key differences and why they matter for marketers
The rise of generative AI in content production, particularly video, has also led to a rise in the risks associated with it. But discriminative AI may be the solution, says Emma Lacey (SVP EMEA, Zefr). For The Drum's Web3 to AI Deep Dive, Lacey looks at the key difference between generative and discriminative AI, and why the latter may be the safest bet for marketers.
15% of UK businesses are now adopting at least one form of AI technology - but it's crucial to tread carefully
Excitement around generative artificial intelligence (AI) is growing – not least because of its potential use when it comes to video. For example, the freshly announced ChatGPT-4 has taken the ‘fastest growing app in history’ through another phase of innovation, bringing with it this time text to video capabilities.
Of course, AI has been rising up the business agenda for some time, with 15% of UK businesses now adopting at least one form of AI technology. But while AI tools are gaining ground, it’s essential to acknowledge that smart tech implementation is still in its very early stages, meaning it remains key to approach with caution and keep safeguards in place.
Although generative AI has stolen the spotlight of late, increasing interest has also underscored the need for other intelligent solutions to help manage automated content effectively, like discriminative AI. The rise of generated content is nascent, and the increased risk of misinformation it presents should encourage advertisers to be mindful of how and where they place ads.
What is generative AI and why do content producers want to use it?
Generative AI produces content using deep learning algorithms, rather than analyzing or acting upon existing data. Generative AI can create a wide range of content, from written text to images and now, potentially, video, and tools are already disrupting a wide range of industries; from entertainment, art and design, to advertising and media. With AI technology such as ChatGPT able to generate content from user prompts in a matter of minutes, it’s no surprise that publishers might see this as an opportunity to streamline content production.
For media companies and brands hoping to increase their output, the appeal of generative AI tools is highly compelling due to the faster and cheaper production at scale. For example, Forbes has its own machine learning AI tool, Bertie, which it uses to generate descriptions, optimize headlines and recommend images to save time for its journalists. However, as with any developing technology, there are still limitations in the use of generative AI tools, and relying on them too heavily can create challenges, particularly when it comes to the more complex area of digital video advertising.
What risks do advertisers face when placing ads beside generated content?
As auto-generated content is on the rise, so is the threat of misinformation. Generative AI tools create content based on user queries and ‘learned’ patterns of language. This ‘learning’ is web-based – so the information may be unchecked, biased or irrelevant – which can easily lead to misleading and inaccurate content. Misinformation is currently considered to be the 16th highest risk to global society, and in a climate where it only takes one mishap for brands to tarnish their reputation and damage their consumer relationships, this presents serious concerns.
Placing ads besides AI-generated content could see brands accidentally associated with content that contradicts their values, undermines their mission or may even be dangerous. In fact, OpenAI has warned that the latest version of GPT remains unreliable and may ‘hallucinate’, meaning it may invent facts or make reasoning errors. The lack of nuance in current generative AI tools is something for brands to be aware of to protect their image and reputation. One way advertisers can minimize this risk and avoid unintentional misplacement of ads is to look to discriminative AI to determine the most suitable content to advertise alongside.
What is discriminative AI and how can it help advertisers?
Discriminative AI uses additional, human-generated insight to inform the assessment of online content, evaluating its validity and relevance for advertisers. Its core safeguard stems from leveraging human intervention rather than relying solely on machine learning. By using clearly defined content categorizations, it is possible to conduct real-time semantic analysis against unique preferences, leading to a clearer understanding of the suitability of any ad placement.
Using discriminative AI for ad serving is a sophisticated and nuanced way to reduce the risk of misplaced ads and alignment with harmful misinformation, while directing ads towards more suitable content. Another advantage for advertisers is that the ad targeting informed by human-generated insights is more likely to engage key audiences, and provide further access to like-minded consumers.
As video content is much harder to monitor than text, and as AI tool capabilities turn towards video, advertisers are sure to find themselves taking a more nuanced approach to placing ads. Enthusiasm about the possibilities of auto-generated content is immense, and rightfully so, which is why remaining ahead of the curve is a priority. While AI accelerated production might expand opportunities for advertising reach, utilizing discriminative AI where possible to avoid the risks of advertising against generated content is a key.
Content by The Drum Network member:
Zefr is the most sophisticated platform for walled garden investment.Find out more