In generative AI’s moment, prompting basking in its glory

When Andrej Karpathy, the former head of AI at Tesla who now works for OpenAI, tweeted “the hottest new programming language is English” a few months back, it was prescience at its best.

With AI platforms like ChatGPT all the rage, knowing how to optimise responses is the next big challenge, resulting in a whole new universe where it’s all about phrasing that perfect prompt. One can say that if generative AI is having its moment of glory, so is prompting.

Also read: Twitter’s head of trust and safety says she has resigned

A host of platforms and features promise that perfect prompt, vacancies are being advertised for prompt experts, there are prompt marketplaces, prompt engineering is a buzzword of sorts, and online courses abound on how to become one.

As Google CEO Sunder Pichai said at the company’s recent I/O 2023 event, “While AI is having a very busy year, it boils down to the prompts given to the AI chatbots, image and video generation tools.”

Google introduced a feature called Sidekick at the event for exactly this purpose. Putting the right prompt “can be daunting for many of us to even know where to start. Well, what if we can solve that for you,” Aparna Pappu, Google’s VP of engineering, said at the event. Sidekick offers prompts based on the Workspace doc a user is working on. The live demonstration Pappu offered was that of a children’s story in progress, where Sidekick offered prompts on what could happen next.

There are several prompt generators available online. Launched in late February this year, PromptPerfect is one of them, meant to “elevate your prompts to perfection” for tools like ChatGPT, GPT-4, Claude, DALL-E 2, and Stable Diffusion. The tool has so far helped “94,335 users make 2,095,413 better prompts. That is 909 prompts per hour,” as per the company’s website. Other similar prompt generator platforms include AI Text Prompt Generator for image generation tools like Midjourney and Stable Diffusion, PromptoMania, PromptBase, etc.

PromptBase is a prompt marketplace where you can not only find prompts, but sell them too. “You can sell your own prompts and start earning from your prompt engineering skills,” the website says. Notably, $2.99 is the minimum at which prompts are selling on the platform.

Also read: Samsung launches new OLED TVs with Neural Quantum Processor 4K: Details

A sum of $280,000-$375,000 is what you can expect to earn if hired as a prompt engineer, as advertised by California-based AI startup Anthropic. Although a new ‘field’, a brief scroll on LinkedIn can reveal the interest to hire for such a role.

“With any technological innovations comes the need for skilled professionals. AI prompt engineers have become one such necessity with an increased focus on generative AI. Many companies, including in India and globally, have started offering jobs to prompt engineers to design precise instructions for AI chatbots and carry out tasks like generating code suggestions, identifying coding errors, improving coding performance, troubleshooting and debugging code,” said Prashanth Kaddi, partner, consulting at Deloitte India.

Of course, the internet has help how to become a prompt engineer too, with several online courses claiming to teach that. One such course is being offered by DeepLearning.AI in partnership with OpenAI. Andrew Ng, DeepLearning.AI founder and co-founder of Coursera, and Isa Fulford, a member of OpenAI’s technical staff are the instructors. A similar but lengthier course, where one can “enroll for free” is also available on Coursera, for which over 19,746 people are already enrolled. Online learning platform Udemy, too, hosts a number of such courses, one of which is 15-hour-long, with over 13,000 students enrolled.

Kishan Panpalia, a member of the founding team of AI-driven content marketing platform Pepper Content, which was “one of the earliest adopters of GPT-3”, says the role is here to stay. “If generative AI continues to be commoditised and democratised as it is being done now, there will be a tonne of products that will be made as used cases. Hence, brands will either have to take generative AI as a service from outside or develop in-house. In either case, they will need prompt engineers.”