OpenAI’s large multimodal model GPT-4 can do it all: interact with text and images, write poems, and solve complicated math problems. This will likely have professionals all over the world wondering: will my competency still be relevant a few years from now?
OpenAI has released the latest version of its large-language AI model, GPT-4, making headlines with suggestions of improved capabilities. Although its engineers disclosed that the bot is not perfect, in the US it managed to score 90% on the bar exams and a near-perfect score on the high-school SAT math test.
In addition, GPT-4 is much more advanced than its predecessor ChatGPT: it can engage with visual inputs as well as textual prompts and is capable of completing tasks based on pictures and drawings. For comparison, GPT-4 can process over 64,000 words as against GPT-3.5’s maximum of 8,000.
This caused much concern amongst employees already stressing about accelerating automation and job cuts at major companies. We asked experts for their opinion on whether GPT-4 is that much of a threat to human workers – and which jobs will likely be lost to the powerful chatbot.
Writers, marketers, designers… watch out
Already in February, anecdotal reports saying that ChatGPT can make anyone a millionaire surged, with one TikTok user offering to charge companies up to $1,000 for a copy produced by ChatGPT instead of a real writer.
That is not amusing, according to Max Thake, co-founder at peaq, which bills itself as “a Web3 network powering the Economy of Things.” He’s expecting writers to be among the first to go in the new GPT-4 world.
“It’s hard to immediately assess the new model’s capabilities, but from the looks of it, the list is pretty long. GPT-4 is about to affect jobs for coders, writers, marketers, designers, both web and visual – as it can generate prompts for other generative models like Stable Diffusion. In my view, it will be entry- and medium-level jobs at startups and SMBs [small and medium businesses] that will take a beating.”
Thake then moves on to the marketing industry as an example – startups focused on saving costs may leverage GPT-4 to create and execute commercial strategies using the bot directly. The same applies to coding, where a business can have GPT-4 do a basic task or use it to build a simple website. In this sense, the bot eliminates the need for a middleman in areas where it can match – or even outperform – a human.
“Also, integrating generative AI, including GPT, with various connected devices can effectively transform the way we interact with technology, wiping off [sic] a host of manual jobs,” he said.
“For example, how many movie studios would pay a drone operator if they can [directly] tell the drone what to do? How many cafes would opt for a robot barista that can recommend a drink based on your mood? For now, GPT and its likes may seem simply exciting, but in a few years, it will have transformed entire industries.”
A nuanced problem
Now of course, the question of jobs being taken by GPT-4 gets much more nuanced when other details are taken into account. For example, does the seniority of a role matter, or will every customer service position across every industry get eliminated?
The possible issue of machine hallucination should be taken into account, with OpenAI’s CEO Sam Altman tweeting about GPT-4’s flaws, saying that it “still seems more impressive on first use than it does after you spend more time with it.”
Similarly, Richard Gardner, CEO of Texas-based Anthropic AI, offers a different take, suggesting that not all professions are equal before GPT-4, and that the level of expertise and creativity required for a project will determine the outcome.
“The jobs most likely to be automated first are those which are repetitive and routine, such as data entry and customer service. Those involved in more complex jobs are more likely to use GPT-4 as a tool, rather than see it as a replacement.
“Entry-level writers who simply ‘spin’ content to create massive amounts of text are much more likely to see GPT-4 as a threat than high-level creative and technical writers, for example, who will likely utilize the technology as an aid to complete their projects more quickly. Salespeople, for example, could see extreme value in large-language models for their work. It could be the difference between turning great sales teams into phenomenal sales teams.”
Thake echoes that view, saying that while GPT-3 was already advanced enough to replace some workers, it would all depend on the desired quality of the final product.
For example, although GPT-3 was capable of producing content, its writing was far from sophisticated. And while Thake expects GPT-4 to do better in that regard, he highlights that GPT-3 could not replace a professional writer, but rather helped with generating ideas.
OpenAI seems not only to understand that, but to highlight it too. The MIT Technology Review has pointed out that the release of GPT-4 was treated more like a product tease than a research update.
While GPT-3 was groundbreaking during its launch, GPT-4 cannot quite reach the same level of popularity, especially when faced with severe competition from other large-language machines such as Visual Language Models Flamingo from DeepMind.
Ultimately, GPT-4 still has a lot to learn, battling its way through hallucinations, bias, and hate speech – and teaching a bot ethics will likely take some time.
Learning to live – and work – with the machines
GPT-4 is already here, and it’s time for us to learn to live with it. But before freelancers and workers start considering a new career path, it’s worth remembering that we’ve been here before – with the invention of the internet.
“Although GPT-4 is a big advance, in some cases, compared to GPT-3.5 the question of jobs is more nuanced. This could be summarized in the phrase we see again and again on socials about this topic: ‘AI will not take your job, but someone using AI will.’
“It means that AI will be part of our life, like many other technological evolutions in the last few decades, so the real danger is staying behind in embracing it to boost productivity.
“Phones did not replace in-person meetings, X-ray machines did not replace doctors, and calculators or computers did not replace mathematicians. They all enhanced and extended the capacity of the people,” says Giancarlo Erra, founder and CEO of AI-based startup Words.Tel.
After all, GPT is a natural-language-processing model, meaning that it takes cues from existing data: it is not capable of producing creative ideas the way humans can.
However, GPT-4 can certainly come in handy both for organizations and workers – but as a support tool rather than an actionable replacement. The government of Iceland, for example, is collaborating with OpenAI in its GPT-4 project to preserve its language. This could also serve as a basis for a bigger initiative aimed at promoting the preservation of other endangered tongues.
The visually impaired service Be My Eyes utilizes GPT-4 to provide blind and partially sighted people with information previously inaccessible to them. Morgan Stanley uses the model as an internal research tool for its staff of financial advisers, training it on 60,000 reports on the global economy and 40,000 internal documents from the firm.
There is no doubt that similar language models will transform the way we work, changing the operations and allowing employees to occupy more valuable or creative roles. But while companies should outsource manual planning and simple content production to GPT-4, expert tasks should always be reserved for humans.
“The average level and reliability of the result of AI with human intervention/filter usually need improvement. Images generated with AI are only good enough with re-touching or with proper prompt research and the addition of human work,” said Erra.
“Content created by AI can be wildly inaccurate, written in a correct but very 'mechanical' way, and missing the thinking that results from understanding the topic. With development, the code usually works, but things like security, future-proofing, and performance still need to be included. Also, the code is often old and needs adjustment.”
Erra’s final verdict, for now, is that “AI is an invaluable integration tool but won't directly replace jobs.” However, he added: “Workers and companies not embracing it could be left behind.”
More from Cybernews:
Subscribe to our newsletter