When OpenAI released ChatGPT-3 in November, especially after announcing the GPT-4 version in March, there was a flurry of articles about prompt engineering in the mainstream media and business press.
“AI ‘instant engineer’ jobs can pay up to $375,000 a year and don’t necessarily require a tech background,” he promised in an Insider article in April. Time and Forbes magazines also joined the chorus.
But prompt engineering—creating queries for generative AI that reveal more sophisticated and informative results and help train tools—is unlikely to be a long-term career path, according to AI organization leaders. Low.
HackerRank co-founder and CEO Vivek Ravisankar said, “This is a short-term fad and it will go away. You just start learning how,” he said. He is using AI to create skill-based hiring reviews for engineers, he told The New Stack.
“Interfacing with AI will be a real skill, much like interfacing with Google as a search engine.”
How else will generative AI change the technology job market? In a survey released in April by HackerRank, 82% of participants believed generative AI would redefine the nature of coding. Did.
The truth is, no one knows yet what impact a fast iteration tool like ChatGPT will have on the tech job market. However, some clues have already emerged.
Tribe AI is a 4-year-old consulting company that recruits AI talent and matches it with customer projects. According to Tribe AI’s Jacqueline Rice-Nelson, customer inquiries have increased 100% month-over-month this year, “exploding” demand. Co-founder and CEO.
“The market has completely changed,” Nelson told Newstack. “It’s not just how companies perceive what is possible, but how it perceives that it is not only possible, but important to act. It takes everyone from “one day we might be an AI company too” to “we might be an AI company”. offal To become a company that can utilize AI today. And because these tools exist, it is possible. ”
Fortunately, Tribe AI has also seen a significant increase in the number of people applying to join its network of AI talent, with a 50% monthly increase in 2023, Nelson said. And the quality of candidates has improved compared to before the introduction of GPT-3.
“Interfacing with AI will be a real skill, much like interfacing with Google as a search engine.”
— Vivek Ravisankar, HackerRank
She said her company was founded because she determined there was an imbalance between the demand for AI talent and the supply of AI engineers. “We figured out how to activate talent that didn’t exist in the job market because they didn’t want full-time jobs, they wanted fractional jobs.”
In response to this demand, Tribe AI was able to gather hundreds of members and match them to AI and machine learning projects. Nelson continues to believe that there are more engineers and other experts ready for the next wave of AI. He “feels like we’re not quite there to tap into the true depth of the supply and talent market.”
“Prompt hackers” and AI strategists
As the generative AI revolution takes off, experts say there will be a keen demand for people with deep experience in large-scale language models (LLMs) trained with ChatGPT and other tools.
“If you are an engineer who has trained a large language model of over 10 billion parameters, congratulations, you hit the jackpot,” said Noah Gale, co-founder of Tribe AI. “You are in great demand.”
Nelson said one reason for the demand is that there are very few people with that kind of depth of experience. “I think LLM engineers are probably the hardest to find because a lot of them are still in places like Google and OpenAI,” and so on. ”
But companies need rapid change, Nelson said.
“It’s a completely different project,” she said, but that was before ChatGPT-3 was introduced. Now, she said, companies are asking, “What should my generative AI strategy be?” I have an idea that could use generative AI, does it make sense? Is that where I should focus? ”
That direction requires “someone with a deep knowledge of how these models work and what you can do with them—the power of models,” she said. rice field. But you also need a sense of product and strategy. An engineer who can speak English. And that’s true even for less technical people. So I think that changes the profile to be more of an AI strategist. ”
The best candidates for some generative AI roles may have more in common with growth marketers than with engineers, she said. “Some of the very good profiles we’ve seen in this field have been completely counter-intuitive. For example, improv comedians. He’s very good at coming up with things on the spot.”
Gale said there is a growing demand for “nimble hackers” at Tribe AI — hackers who are capable of breaking through the base model to expose vulnerabilities in order to improve the model.
“This is very important for these foundational model companies to make sure what they are actually putting out there is safe,” he said.
Gale said Tribe AI has seen interest from organizations seeking help fine-tuning generative AI tools. He said that once organizations start using these tools, the natural next step is to train models on their own internal data, perhaps building entirely new models specifically for their use cases. It is said that it will be
Dylan Fox, founder and CEO of AssemblyAI, also advocated training a generative AI on an organization’s own data (“making it aware of all the company’s internal documents and Slack histories”). ”) said AI-related jobs could be a growing area in the future. Transcribe and analyze audio data.
Gale offered what he called “very hot opinions” on the near-term future of Prompt Engineering-related roles, stating, “Within a year, there will be people making $1 million or more in this role. And within a year of that, the role would cease to exist.
“These large-scale models get so good through reinforcement learning and training on our own data that they no longer need to understand our intentions. [us] You can be very specific and creative with these prompts. it will just tell. ”
Recruitment of system engineers
But it’s not just AI/machine learning engineers and “smart hackers” that will be in high demand, experts say. UX and full-stack engineers will also be hotly sought after by organizations looking to use new generative AI tools.
Nelson says these are “complementary roles built around advanced AI technology, because you really need to know where to put things.” And we need to incorporate that into our products in a way that allows the consumer to engage. And it should fit the backend.
“It’s kind of the missing piece in the story. [for] salt stack. ”
Operations and systems engineers may be the unsung heroes of the generative AI revolution, experts tell The New Stack.
“A lot of what people are doing with LLM today is abusing the laws of architecture and scaling and designing larger and larger data centers to train larger and larger models. and requires a lot of engineering work.”
— Dylan Fox, AssemblyAI
AssemblyAI’s Fox marveled at the systems engineering required to get generative AI to where it is today, saying, “It’s more an engineering achievement than a research achievement.”
he asked: “How can we stitch together hundreds of servers with GPUs, communicate with each other with very low latency, move large amounts of data with very low latency, and work very efficiently? ‘Trying to find a good model?
“A lot of what people are doing with LLM today is just abusing the laws of architecture and scaling and designing larger and larger data centers to train larger and larger models.” He said. “It’s just scaling and it takes a lot of engineering work.”
Fox’s company is working on what he calls “AI’s Blue Apron.” Mealkit, like his service, offers user-friendly yet customizable products, which he envisions as “a superhuman AI system that delivers multilingual, low-latency, automatic speech recognition.” .
As he grew the company, he insisted on hiring two to three engineers for each AI researcher. “A lot of it is scaling up datasets and training. He is akin to scaling up infrastructure, which requires very deep and complex system engineering.”
Be curious and stay relevant
In a new HackerRank survey of 42,000 developers, three out of four participants said they would adjust their skills to keep up with AI growth.
With generative AI rapidly changing the way software is built, deployed, and managed in the next few years, how will engineers stay in the game?
Good news. Perhaps what first drew you to development and engineering was the trait of curiosity and willingness to tinker and experiment. (In the HackerRank survey, 4 out of 5 of her respondents said they started using her ChatGPT or similar tools.)
“The most important skill is to be very curious,” Gale said. “And we are working to build, experiment and ship very quickly.”
Generative AI amazes us every day with what it can do. Asked about a recent surprise, Fox told The New Stack about how his non-programmer wife used his ChatGPT to create his Chrome plugin in an hour of his time. The work will probably take half a day, he said.
“AI is not replacing humans. It is humans using this as a tool, and it is replacing people who choose not to use this as a tool.”
— Noah Gale, Tribe AI
And models are getting smarter all the time. The Fox company released a new speech recognition model earlier this spring. “When I was playing with it, I threw in a Drake song,” he said. “The transcription of the lyrics was perfect too. And I thought we weren’t even trained in music for this! It’s crazy.”
But no one is talking about generative AI at The New Stack, and the tech job outlook has gone into doomsday mode when it comes to talking about the impact AI will have on engineers.
“AI is not going to replace humans,” Gale said. “It is humans using this as a tool instead of people who choose not to use it as a tool.
He added, “These language models have allowed one engineer to do the work of ten. It’s really people,” he added.
In addition to experimenting with new tools, Fox recommended delving into the published research on them. “Read the actual research paper,” he said. “Usually when you look at research papers, they are much more sober and conservative than what you see on Twitter or in the news.
Ravisankar noted that innovation always takes some jobs away, but creates others. “In general, it’s better to embrace technology. Even if some things don’t really work out, it’s probably better in the long run. I would give the same advice to: How can we incorporate AI into our daily work?
And he pointed out that many innovations over the last 20 to 30 years have made programming more and more people, not less. “Based on this particular paradigm, this is like another toolkit in a developer’s backpack that not only improves developer productivity but also reduces the barrier to entry for new people. You can almost think of it as something to put down. Go in and build.”
“The next decade or so is going to be a very exciting time,” he said.
