Will AI replace humans in tech startups (+ChatGPT examples)
AI seems to be able to do almost everything these days. This got us thinking - can and will AI replace humans in the nearest future in tech startups and companies?
20 February, 2023Since its launch in November 2022, ChatGPT has been on everyone’s radar. From writing articles to generating code, AI seems to be able to do almost everything these days. This got us thinking - can and will AI replace humans in the nearest future in tech startups and companies?
Basically, ChatGPT, a chatbot built on top of a language model, is the Internet’s new toy, and not without reason. You can chat with it, ask it to give you ideas for your essay or letter, or even help you with your work. Yet, it’s not going to replace humans anytime soon, especially in the workforce.
Long story short - it lacks vision and personality and even has some other, far worse, limitations that definitely hamper its process of taking over human jobs. So, let’s take a closer look at why we, humans, are still going to keep doing what we do for quite some time.
How far has AI gotten so far
The idea that AI can replace human jobs is not new, and Artificial Intelligence as a tool has gotten pretty advanced recently in terms of the variety of uses, the accuracy of predictions, and the creative answers.
Current language models, for example, are trained to recognize and learn from every user input, guessing what a person means when asking a question. In this instance, Machine Learning helps computers make decisions based on previously learned information.
Overall, AI has made significant progress in recent years and has already become a key component in many tech startups, from automating routine tasks to helping with decision-making. However, AI's capabilities are still limited compared to human intelligence, and it hasn’t gotten as far as you think it has.
Even with all the progress, the sample AI we used to prove some points in this article, ChatGPT, still produces hilariously wrong answers, which we’ll discuss later on.
AI uses in the workforce
The most obvious and prevalent AI uses in the workforce are the so-called AI-based products that use Artificial Intelligence as a base for their services. For example, YouTube, Netflix, and their recommendation algorithms. The former uses a recommendation engine consisting of two neural networks, and the latter also has its machine-learning-based recommendation system. Safe to say that their help in analyzing customer preferences is tremendous.
Customer support, marketing chatbots, and smart assistants is another excellent utilization of AI technology. AI and chatbots are integrated into websites, messaging apps, and mobile applications to provide 24/7 support and automate routine tasks. Companies use chatbots to improve the customer experience, increase efficiency, and reduce costs.
AI chatbots are also used in the healthcare, finance, and education industries to provide personalized services and information, for example, answering frequently asked questions, guiding customers through purchasing, and providing other customized recommendations.
So far, the most promise AI shows in terms of use cases is in robotics (especially manufacturing) and data-heavy fields. Professions that rely on processing vast loads of information are the first ones that will get a boost from AI.
For instance, Microsoft just integrated ChatGPT to recap Teams meetings, and this startup recently launched the “world’s first robot lawyer” - a tool that uses AI to be the customer’s advocate.
However, AI usage is still fairly limited. Chatbots and other AI-based have the potential to automate specific tasks performed by information workers, but, likely, they will only partially replace human information workers in the future. While AI can handle routine and repetitive tasks, it struggles with jobs that require creativity, empathy, and human judgment.
For example, information workers bring unique value to organizations through their ability to think critically, problem-solve, and make decisions, tasks that are still difficult for AI to perform. Additionally, human information workers can provide a personal touch and build relationships with customers and clients that is difficult to replicate.
Overall, AI-powered tools are now seen as a cost-effective and efficient way for businesses to improve customer engagement and support. Their use will continue to expand even more as the technology improves and more businesses adopt AI solutions.
AI chatbots may also become more advanced, with improved natural language processing and more sophisticated capabilities, allowing them to better understand and respond to user requests.
AI vs. human: will AI replace programmers in startups
Just a couple years ago, the prospect of AI replacing programmers seemed far-fetched. But now we have AI systems like ChatGPT that can often write working, human-like code even for some fairly complex tasks.
However, AI is still far from replacing developers but rather complementing and enhancing their abilities, enabling them to work faster and delivering higher quality software.
Existing programming tools will likely be integrated with AI solutions in the future, which will boost their productivity and automate a lot of code generation. Smart programmers will still be required to solve complicated problems, do code reviews, and design systems architecture to match business requirements.
New AI systems have an impressive understanding of natural language, and they can turn generic descriptions of problems into working programs. In the example below, the AI chatbot helpfully outputs a whole JavaScript program in mere seconds.
If we run the code given by the AI, we see that it both understood the problem prompt and correctly implemented the code. The code is very human-like, easily readable, and even filled with little comments about what it is doing.
In this example, the AI also demonstrates an understanding of the JavaScript standard library by correctly using the canvas API, even though it has not been explicitly trained to program JavaScript.
A human programmer would have to be familiar with JavaScript to complete this task, but the AI has just learned about it by scraping the Internet. Human programmers do not have the ability to read millions of code examples and learn about every API in existence, but an AI can.
Of course, most of the time, programmers modify code rather than create something from scratch. Any system hoping to replace programmers needs to be able to read code and adapt it to new requirements. With ChatGPT, you can make follow-up requests, and it can often correctly implement the requested code changes.
Again, when we run the code given by ChatGPT, we see that it correctly understood and implemented the requested code changes. This is much closer to real programming work, and it is seriously impressive.
But this is also where you run into some severe limitations of the current AI systems. They have a fairly limited memory of the previous conversation, so they can’t handle larger code snippets. Also, the workflow of copy-pasting things between the chatbot and your programming environment is far from ideal. AI should be integrated into programming environments to be useful for serious work. Without an efficient workflow, the usefulness of AI will be minimal.
But even after some AIs have been optimized for programmer productivity, there are still serious concerns about current AIs’ ability to think critically and do more serious problem-solving.
While existing AIs already have impressive code-generation skills, they are still just language models that have a limited understanding of problems or systems. The Internet is full of examples of ChatGPT failing even fairly basic math problems, which is not a good look for someone posing to solve complicated programming problems.
It is also worth noting that AI can be overly confident in its answers, and its output should still be vetted and approved by human programmers. Programming website StackOverflow banned the usage of ChatGPT on their website, citing a high rate of misinformation and nonsensical answers given by the AI.
For the AI to completely replace human programmers, it would also need to be able to search and understand internal documentation. This type of integration into the company's internals creates more maintenance, compliance, and auditing work and might not be worth doing for small companies like startups. Full company-level AI integration might stay an enterprise tier-only feature for a long time and widen the technological gap between small and large companies.
So, the general consensus is that - no, AI and language models cannot replace developers. While AI and language models can assist with specific tasks, such as code generation and bug fixing, they still lack the critical thinking, problem-solving, and creativity that are the essential qualities of a good developer.
Since AI and language models still have many limitations and need to be trained and fine-tuned by human developers, rather than replacing them, AI and language models can serve as tools to help developers work more efficiently and effectively.
Ultimately, AI will not replace developers but rather complement and enhance their abilities, enabling them to tackle more complex and challenging tasks, and delivering higher quality and efficient software.
Will AI replace designers
The same answer still applies. AI can potentially replace designers in routine tasks, but with stuff that requires creativity and a personal touch, human designers still remain unparalleled. It is said that while AI can probably be a good substitute for a current designer in many of the tasks, it cannot replace designers of tomorrow - the job will evolve, and for the better.
For example, in the following case, we used natural language to describe what kind of design we wanted, and ChatGPT happily complied. In UX/UI design, with the help of AI, designers have the potential to try out different designs and layouts very quickly.
Same with the following case of creating a user flow for an app - ChatGPT did a quick outline, giving inspiration for the app design and saving time on routine planning.
But will AI replace graphic designers and UX designers altogether? Not really. You still need humans to decide on what is a good, user-centered design, conduct user research, and build more complex interfaces.
Will AI replace managers
Finally, will AI replace managers in startups as well? While AI could help with the planning portion of job responsibilities, human managers are still needed to handle the inherently human elements of the job, such as checking up on people, prioritizing resources, and maintaining the company spirit. Take a look at how ChatGPT could help with the monthly development roadmap, for example.
AI limitations
Now onto the fun part. Current AI’s limitations can be summarized into the following:
- Language models like ChatGPT refuse to answer certain types of questions, i.e., predictions of the future, personal opinions, or those leading to offensive or harmful answers;
- Limited memory - it can forget earlier parts of the conversation;
- The quality of responses you get is dependent on the prompt’s quality and the training data;
- And lastly, it can be overly and hilariously confident in completely wrong answers, as you can see in the example below.
Accessibility of AI
Even if AI technology that can help in software development and design exists, it is only helpful to startups if they can buy or acquire the AI with a reasonable budget.
With Microsoft recently investing $10 billion in OpenAI and each tech giant like Google and Meta having their own competing AI labs, the future accessibility of AI remains unclear. But undoubtedly, each company with access to AI technology will seek to maximize its profitability.
So far, the general public has had good access to AI technology with things like the free ChatGPT service or open-source Stable Diffusion that many people run on their home computers.
However, there is no guarantee that this will continue, and it is possible that in the future small companies will be at a disadvantage if the big players have AI technology that drastically improves their efficiency.
Task-specific AI
The quality of the answers given by an AI is highly dependent on what the AI has been trained to do. Chatbots like ChatGPT have been trained to provide ‘helpful’ answers to general questions, but this approach limits its usefulness in any specific task.
Just like today, we have humans who specialize in some skills; in the future, we will likely have AIs with specific skill sets that are able to give much more elaborate answers to human-given prompts.
One example of such task-specific AI is Dramatron which can write whole screenplays based on simple one-line descriptions of the film. Dramatron has been tuned to write only one type of output (screenplays), so it is much more useful in its own field than a general chatbot.
It is likely that in the future, a tech company that wants to fully utilize AI in its operations would need to have multiple AI systems, each tuned to help in different functions at the company.
What does AI have to say about it
It’s only fair to ask what AI “thinks” about it. Even though it “doesn’t give personal opinions” on any matter, if asked how soon it thinks the chatbots like itself can replace information workers, for example, then the reply is simple - it is difficult to predict precisely when chatbots will replace information workers, if ever.
Overall, it will depend on several factors, including technological advancements in artificial intelligence, the ability of chatbots to handle tasks that require human skills, and the willingness of organizations to adopt chatbots in place of human workers.
Currently, chatbots are used to automate routine tasks such as answering frequently asked questions, but they cannot still handle more complex tasks that require human skills. Until chatbots can consistently perform tasks that require human judgment, creativity, and empathy, they are unlikely to completely replace human information workers.
Here’s the excerpt from one of the chats we had with ChatGPT.
Why AI can’t replace humans - a summary
To summarize all of the above - no, AI and language models cannot replace humans in tech startups. Not yet, and not for a long, long time.
While AI and language models can assist with specific tasks, such as data analysis and generating reports, they do not possess the creativity, critical thinking, and emotional intelligence that humans bring to the table.
Startups do require a combination of technical and human skills to succeed, and AI can complement and enhance human abilities but cannot replace them.
To add insult to injury, many AIs “deliberately” provide wrong answers due to their limitations as language models. As you’ve witnessed in our previous examples, don’t count too much on AI to give you factual information that is very recent or excel at logical reasoning.
Still, AI is very beneficial to humans. Just customer service alone - with advancements in natural language processing, AI chatbots better understand and respond to customer requests, giving us more time to focus on more important and creative tasks.
So, take a deep breath. AI is not coming for your job. Unless…