Excerpt from the Marketing Mix Interview with Mary Kemp.
What is the state of the industry now?
OpenAI has dominated the market, giving it significant influence. However, the emergence of platforms like DeepSeek has been a game-changer, shaking up the industry. While OpenAI currently has a near-monopoly, I don’t expect that to last, and I hope it doesn’t.
One major concern is that if most businesses rely on a single AI model, it concentrates power in one place, potentially shaping decisions and reinforcing biases. These models inherently contain biases. OpenAI’s model, for example, is largely based on an American, middle-class perspective, which doesn’t reflect the full diversity of global users. People need to be aware of these biases and how they might affect them.
Around one-third of the world’s population doesn’t have internet access, meaning their perspectives are not included in these AI models. This creates a significant skew in the data that AI learns from.
Have you tried DeepSeek?
Yes, I’ve experimented with it a little. Interestingly, if you ask it about Tiananmen Square, it won’t provide a response, and it avoids negative references to figures like Winnie the Pooh. However, DeepSeek represents a major paradigm shift.
Due to embargoes on advanced Nvidia chips, DeepSeek had to innovate with less powerful hardware, finding alternative ways to make its model work. The result is an open-source AI, which is a rarity in the industry. Researchers at Berkeley have already replicated it, suggesting that more disruption is on the horizon.
One key takeaway is that no one can truly predict the future of AI. Experts disagree on almost everything and much depends on how individuals and companies choose to use these tools. The technology itself is neutral. What worries people is the potential for misuse by bad actors.
Where does Claude fit in?
Claude is considered one of the most ethical AI models. Unlike ChatGPT or Microsoft Copilot, it isn’t connected to the internet in real time. Instead, it learns from uploaded data and is then disconnected, reducing risks related to privacy and misinformation.
That said, Claude’s adoption is relatively limited. Businesses often opt for more functional alternatives like ChatGPT.
While many large language models now outperform humans in most tasks, my advice is to choose one platform that suits your needs and master it rather than spreading yourself too thin across multiple tools.
Many specialised AI platforms exist. Do they rely on OpenAI?
Yes, most of them do. Many niche AI tools have OpenAI’s model running in the background, with customised instructions to serve specific industries or tasks.
We are also seeing a wave of AI-driven SaaS tools emerge, although some may be vulnerable to changes in OpenAI. For example, OpenAI’s launch of Sora, a text-to-video model, may disrupt smaller companies currently offering similar services.
One challenge that AI companies face is regulatory concerns, especially in Europe, where copyright laws and other legal frameworks are stricter. Some AI platforms are hedging their bets and slowing European releases, making VPN (Virtual Private Networks that encrypt your data and hide your IP address) usage more appealing.
Do you primarily focus on ChatGPT in your work?
We train across all major AI platforms, so we remain platform-agnostic. However, we are seeing increasing adoption of Google’s Gemini, particularly after the release of Gemini 2, which has a more competitive pricing model.
Microsoft’s Copilot is another key player, though it is essentially a simplified version of ChatGPT. Businesses appreciate its security features since it integrates within Microsoft 365, even if it sacrifices some functionality.
Of all the models, ChatGPT still leads in terms of functionality. Its ability to be customised with pre-programmed prompts unlocks a wealth of potential for businesses.
Interview by Ian Trevett
—–
Read the Full Interview
Take a look at Mary and Gareth’s website for more AI insights.
You can contact Mary on LinkedIn