The launch of generative AI sent many people into a state of FOMO. Yet you don’t need to be a tech expert to become an AI innovator, says Sopra Steria’s manager Kenneth Aastrøm.
Generative AI might be the biggest tech revolution since the Internet itself. But just like when other technologies become popular, there are numerous insecurities, misunderstandings and even scepticism around the use, and potential use, of AI. One of the most significant is that you need to be a techn wizard to innovate with AI.
You don’t.
The tech job has already been done for you – that’s the beauty of AI. You just need to understand how to make it work for you to create the solution you want. Here’s how:
1. Prompts are the language of AI collaboration
Prompts bridge the gap between technology experts, designers, and other stakeholders, enabling effective communication with generative AI systems. Our experience confirms that 80% of a successful AI project lies in carefully crafting and orchestrating prompts. Meticulous prompt engineering directly influences the quality of the output; invest in understanding the nuances of prompting and you've already paved the way to exceptional results.
Unlock the power of AI – prompt engineering is for everyone.
“80% of a successful AI project lies in carefully crafting and orchestrating prompts (...) Invest in understanding the nuances of prompting and you've already paved the way to exceptional results.”
2. Explore the power of AI alternatives
OpenAI and Midjourney are powerful, but your project's success depends on finding the right AI solution for your specific needs. ChatGPT (or GPT-4-turbo) may excel in many areas, but consider alternatives like Google Gemini if local anguage output is crucial.
For image generation with text, Ideogram might be a better choice than Dall-E. And, don't overlook the cost-effective benefits of GPT-3.5-turbo if you don't require the full power of GPT-4-turbo.
Remember, the world of AI is evolving rapidly. The best models for specific tasks today will be surpassed by new developments tomorrow. Stay curious, explore new advancements, and don't be afraid to let your project take unexpected but inspiring turns.
3. Start small and keep it simple
AI's rapid evolution demands adaptability. Don't invest heavily in complex, customised solutions early on; they may quickly become outdated or available as off-the-shelf products you can purchase more cheaply.
With AI research moving so quickly, tomorrow's solutions might solve today's problems in a flash. Prioritise essential functionality and design for easy updates as language models advance. Avoid reinventing the wheel, unless the journey of building it yourself is the reward. Solutions from OpenAI, Apple, and others may soon address the specific problems you're trying to tackle.
4. Harness your data for custom AI experiences
Including your own text, images, or documents in prompts is a quick way to personalise AI responses. However, remember that AI models have a "context window". This is a limit to how much information they can process at once so avoid overwhelming it with the complete works of Shakespeare,
If you need more capacity then try the GPT Assistant. Create a GPT Assistant in the OpenAI web UI to upload documents and other data. This provides more storage and a wider context window for the AI to draw from, unlocking enhanced personalisation.
Hitting the context limit? Explore GPT Assistant or RAG for more robust AI solutions.
5. Be transparent
Following the EU AI Act, be proud that you're using AI in new and innovative ways. Let your users know when they're interacting with AI-generated content. This fosters responsible AI use, promotes open communication, and aligns with fundamental human rights principles such as fairness and non-discrimination. AI is a powerful tool, but it shouldn't be used to judge people – human oversight remains crucial.
6. Be mindful with privacy and data protection
Respect privacy laws and regulations, including GDPR. While sensitive data might not be ideal for cloud-based AI services, explore open datasets or AI-generated synthetic data for experimentation.
For maximum security and control over sensitive information, consider open-source language models like LLaMA for on-device processing. Powerful handheld devices with neural engines like the recently announced iPad Pro's M4 chip are opening up new possibilities for local AI processing.
In summary getting started using artificial intelligence to create new solutions doesn’t need to be complicated. Think simple, focus on gathering good quality input and be clear as to what the purpose is. Let technology solve the rest.