The Risks, Rewards & Best Practices of Generative AI
Let’s start with the basics. Generative AI, the biggest ‘buzz word’ in the Artificial Intelligence (AI) space, refers to the process of using intelligence to generate new data, as opposed to analyzing and categorizing existing data – which is a more traditional AI practice. Put simply, Generative AI is a type of artificial intelligence system or algorithm that is capable of generating new content; including text, images and media in response to a set of prompts or inputs.
As so often happens when a shiny new toy arrives in the digital marketing playpen, we all want to get a piece of the pie, and for good reason, but Generative AI is not without its risks. While an invaluable tool in your roster, there’s a number of things you need to take into account before diving straight in.
Let’s start with the positives.
The Rewards
The biggest perk of Generative AI is undoubtedly the increase in operational efficiency it generates (or, can generate). Automating tasks that would otherwise need manual involvement, Generative AI can help you save a significant amount of time, and resources by automating mundane tasks. For example:
Keyword Brainstorming – Generative AI is a useful asset when in the early stages of keyword and topic research. An AI algorithm can be used to analyze search data and generate a list of relevant keywords – in line with the initial query – that fall into both popular and low competition categories.
Running A/B Tests – AI can be used for A/B testing, but only in relation to the content creation side of things – i.e. generating the text for each of the tests. It’s worth noting that Generative AI tools often fail when asked to conduct the simplest of analytical calculations so don’t use it to judge test outcomes, test validity, statistical significance or anything similar.
Image Generation – The simplest use of generative AI is the access businesses now have to an infinite library of images. These can be used (for now) commercially, within marketing campaigns and so on without risk of copyright infringement.
Connected Asset Generation – Connecting an ever-updating language model with generative video and speech capabilities will allow for an infinite volume of video assets that can be tailored to individual consumers or use cases. Video and other content won’t just be dynamic, it’ll be tailored to any use case for any business.
Take a look at our ‘30 Ways to Use AI in Digital Marketing’ whitepaper for more reasons why you should integrate intelligent solutions into your digital strategy…
The Risks
As we all know, with reward comes risk, and similar to the benefit it reaps, there are also a number of risks that you need to take into account when it comes to Generative AI, including:
False or Incorrect Information – Due to the lack of a singular, clear information source, there is a significant risk of misinformation and/or synthesized disinformation. Generative AI platforms are known to regularly generate false responses, known as hallucinations, or used to produce ‘deep fakes’ that are becoming increasingly harder to identify. (Always, always double check).
Privacy – One of the main concerns surrounding Generative AI is the lack of privacy legislation to protect user data. There are several ways this can impact users, namely; consent violations, inadequate anonymization and the unauthorized storage, processing and sharing of data. There is no current governance/protection regarding confidential or sensitive information. Users should assume that any prompts entered into a Generative AI tool will become public information.
Bias and Discrimination – Although most tools have precautionary measures to mitigate the risk of bias and or discrimination, it is far from 100% fool proof. Even the most enhanced Generative AI can, and is, inherently hindered by the data it’s trained on. Be that bias, racism, sexism and so on.
Weaponized AI – AI can be manipulated to be used with malicious, and/or destructive intent – namely to overcome cybersecurity measures and commit fraud, theft or money laundering (to name a few). Cybercriminals tend to target those with vulnerabilities in their cybersecurity, using malicious AI to bypass ‘benign’ AI – often reverse engineering the functionality for malicious purposes.
Yes, ChatGPT is amazing and impressive. No, @OpenAI has not come close to addressing the problem of bias. Filters appear to be bypassed with simple tricks, and superficially masked.
And what is lurking inside is egregious. @Abebab @sama
tw racism, sexism. pic.twitter.com/V4fw1fY9dY— steven t. piantadosi (@spiantado) December 4, 2022
Best Practices…
Taking the known risks into consideration, there are a number of best practices brands should follow in order to safeguard against malicious or harmful Generative AI usage.
1) Don’t feed your organization’s sensitive/intellectual data into a Generative AI tool as a prompt. All inputted data is subsequently used to further train the AI models, meaning that the sensitive or disclosing prompts entered could be revealed to others using the tool if the query is relevant. We don’t want a repeat of Samsung’s coding fiasco – Samsung workers made a major error by using ChatGPT…
2) Don’t rely on hallucinations or take actions on them – although they may seem convincing, they’re falsified and could have serious consequences for you and your organization. With this in mind, do set out organizational guidelines documenting the policy if someone does breach AI best practices and uses hallucinations to make business decisions.
3) Double check everything. As a whole, you should never rely solely on the output of Generative AI, use your common sense and check all information against reality.
4) If investing in Generative AI solutions, be mindful of the sustainability risks. Significant volumes of electricity are used in AI, so choose carefully when selecting your vendor, and consider documenting the carbon footprint generated via your organizational AI usage and reflect this within offsetting initiatives.
5) AI is only as good as the data that’s fed into it – a quality in, quality out mentality. Be sure to focus on quality, richness and connectivity to maximize your ROI (continually updating as you go).
6) Consider leveraging the power of bespoke AI solutions to maximize the impact of your available budget and discover new opportunities to further drive growth – across Media, Creative and Experience. Great performance is a result of seamless cross-channel integration, and intelligent solutions, tailored to your brand can help you achieve it.
7) Prepare your organization for AI led cybersecurity and fraud risks. Make sure your employees are aware of the risks, and that you have mitigating controls in place if there is a breach of security.
8) Proceed with caution and introduce an internal Generative AI usage policy to mitigate the risks of misuse – with listed examples of acceptable, and non-acceptable uses of Generative AI within the workplace.
For more information on how you can use Generative AI within your organization, get in touch with the team today or check out some of our latest AI based content to really get the brain working:
Browse: Industry Insight
Read Next
Find out how we can help you
With offices around the world, we can build a team perfect for your needs.
