The frenzy surrounding the launch of Large Language Models (LLM) and other types of generative AI (GenAI) is not waning any time soon. GenAI users discover and recommend new and interesting use cases for business and personal life. Many recommendations start with the premise that GenAI requires human prompting. that’s right, time The magazine recently declared “prompt engineering” to be the next most popular job, with salaries approaching $335,000. Tech forums and educational websites focus on rapid engineering, Udemy already offers courses on the subject, and several organizations we work with are now learning how to best use ChatGPT. We are beginning to invest significant resources in training our employees on
However, it might be worth pausing to consider other ways to interact with GPT technology that will likely come soon. We combine our own research into the development of GenAI with conversations with companies looking to develop these versions to show you how to intuitively think about the issue.
Framework for GPT interactions
A good starting point is to distinguish who is who. involved in interaction — an individual, a group of people, or another machine — and who initiates the interaction, human or machine? This allows for 6 different types of his GenAI use shown below. ChatGPT, where one human initiates a conversation with a machine, is already well-known. We then describe each of his other GPTs and outline their possibilities.
Coach GPT is a personal assistant that offers a set of suggestions for managing your daily life. These suggestions are based on user behavior and observations of the environment rather than explicit prompts from the user. For example, we may observe you as an executive and find it difficult to build trust within your team. It can then recommend precise actions to overcome this blind spot. We may also be able to offer personalized advice on development options and salary negotiations.
CoachGPT then sees which recommendations you have taken, which you have not taken, which ones have benefited you and which ones have not improved your advice over time. Over time, you will have highly personalized AI advisors, coaches, or consultants.
Organizations can employ CoachGPT to advise customers on how to use their products. Whether it’s a construction company providing CoachGPT to advise end-users on how to best use equipment, or an accounting firm providing real-time advice on how best to complete a chain of accounts. transaction.
For CoachGPT to be effective, individuals and organizations must enable it to work in the background and monitor their online and offline activities. Clearly, we need to take privacy seriously before we commit our innermost thoughts to systems. But the potential for positive results in both your personal life and your work is immense.
Group GPT They may be bona fide group members who can observe interactions between group members and contribute to the discussion. For example, doing fact checks, providing an overview of the conversation, suggesting what to discuss next, acting as the devil’s advocate, offering a competitor’s perspective, stressing ideas You can even run tests or suggest creative solutions to problems. hand.
The request may come from an individual group member, or from a team manager who does not need to participate in the team’s dialogue, but simply for the purpose of managing, motivating and evaluating group members There is also Contributions can be tailored to the person’s role, skills, or personality and delivered to the entire group or to specific individuals.
The above privacy concerns apply to GroupGPT, but if addressed, organizations could benefit from using GroupGPT for project management, especially in long and complex projects involving relatively large teams across different departments and geographies. , you may be able to leverage GroupGPT. GroupGPT overcomes human limitations on information storage and processing power, making it ideal for supporting complex and distributed teams.
BossGPT Takes an active role in advising people in the group on what they can or should do without being prompted. You can offer individual recommendations to group members, but the real value is to start coordinating the work of group members and dictate who should do what as a group to maximize team performance. sometimes appear. BossGPT can also step in to provide personalized coaching and further recommendations as project and team dynamics evolve.
The algorithms required for BossGPT to work are much more complex, as they have to account for somewhat unpredictable individual and group reactions to commands from machines, but they have the potential to be used in a wide range of applications. I have. For example, if an executive changes jobs, she can request a copy of her reaction to her BossGPT directive in the original organization. This can be used to assess how it fits into the new organization, and her BossGPT specific to the new organization.
At the organizational level, companies can adopt BossGPT to manage their employees, thereby augmenting or even replacing existing managers. Similarly, BossGPT is highly applicable in coordinating work across an organization and managing complex supply chains and multiple suppliers.
Companies could commercialize BossGPT and provide AI solutions to help customers manage their businesses. These solutions can be natural extensions of the CoachGPT example discussed earlier. For example, a company that sells construction equipment can offer BossGPT to coordinate many end users on a construction site. Accounting firms can provide BossGPT to coordinate the work of a customer’s many employees to perform accounting functions in the most efficient manner.
AutoGPT This involves a human giving a request or prompt to one machine, which then engages other machines to complete the task. In its simplest form, a human tells a machine to complete a task, but the machine realizes that it lacks certain software to do it, so it downloads the missing software and He Googles it before installing and using it. Finish the request.
In a more complex version, a human could give AutoGPT a goal (such as creating the best viral YouTube video) and tell it to interact with another GenAI to iteratively come up with the best ChatGPT prompts to achieve the goal. increase. Machines then begin the process by suggesting prompts to another machine, assessing the results, and adjusting the prompts to get closer and closer to the end goal.
In its most complex version, AutoGPT can take advantage of other GPT features listed above. For example, team leaders can assign tasks to machines to maximize both team member efficiency and job satisfaction. AutoGPT switches between individual coaching via CoachGPT and offers suggestions for smoother team interactions via GroupGPT, as well as specific instructions on what needs to be done via BossGPT. can be issued. AutoGPT can then collect feedback from each activity and adjust all other activities to achieve the specified goals.
Unlike the version above, which is still in the making, a version of AutoGPT was developed, published in April 2023, and rapidly gaining wide acceptance. The technology is still not perfect and needs improvement, but it is already clear that AutoGPT can complete a series of jobs that require several tasks to be completed one after the other.
Its greatest application can be found not only in complex tasks such as supply chain coordination, but also in areas such as cybersecurity. For example, an organization can encourage his AutoGPT to continuously address cybersecurity vulnerabilities. In that case, you should look for that vulnerability, which has already happened, but AutoGPT doesn’t just flag threats, it searches for solutions to threats, or finds its own patches to combat threats. You will have to create A human may still be in the loop, but AutoGPT’s response is likely to be faster and more efficient as the system self-generates within these limits.
Imperial GPT GenAI is the most abstract and perhaps the most transformative GenAI, where two or more machines interact, direct each other, and ultimately direct humans through a course of action. This type of his GPT worries most AI analysts who are afraid of losing control of the AI and of it “going rogue”. We agree with these concerns, especially when there are no rigid guardrails to AI execution as it is today.
At the same time, when ImperialGPT is allowed to come up with ideas and share them with humans, but is limited in its ability to act on those ideas, it can be very We believe there is potential for interesting creative solutions to emerge. Lack of knowledge and creativity. You can then easily imagine and execute multiple black swan events and worst-case scenarios with potential costs and consequences to provide possible solutions.
Given the potential dangers of ImperialGPT and the need for stricter regulation, we believe that it will be slow to spread, at least commercially. However, it is expected that governments, intelligence agencies and the military will be interested in deploying his ImperialGPT under strictly controlled conditions.
Impact on your business
So what does our framework mean for businesses and organizations around the world? First and foremost, take a step back and think of ChatGPT’s recent advances as just the first application of new AI technology. is recommended. Then we encourage you to think about the various applications outlined here and use our framework to develop applications for your own company or organization. Along the way, you’ll discover new types of GPT not mentioned before. Third, we recommend classifying these various GPTs in terms of potential value to business and development costs.
We believe that applications that start with a single person initiating or joining a conversation (GroupGPT, CoachGPT) are probably the easiest to build and should generate significant business value, making them a good first choice. is. In contrast, applications involving interactions involving multiple entities or machine-initiated interactions (AutoGPT, BossGPT, ImperialGPT) can be more difficult to implement due to more complex ethical and legal implications. There is a nature.
You can also start thinking about the complex ethical, legal and regulatory concerns that arise with each GPT type. Failure to do so exposes you and your company to legal liability and, perhaps more importantly, unintended harm to humanity.
The following recommendations depend on the type of company. Technology companies and startups, or companies that have sufficient resources to invest in these technologies, should start working on developing one or more of his GPTs described above. This is clearly a high-risk, high-return strategy.
In contrast, if GenAI is uncompetitive or lacks resources, it may be better to adopt a “wait and see” approach. This means you’ll be slow to adopt current technology, but won’t waste valuable resources on what may only be an interim version of your product. Instead, you can start preparing your organization to embrace these new GPTs by preparing your internal systems to better capture and store data, both in terms of work processes and culture.
The launch and rapid adoption of GenAI is rightfully considered the next level in the evolution of AI and a potentially milestone moment for humanity as a whole. GenAI represents a breakthrough in solving fundamental engineering and computer science problems, but it does not automatically guarantee value creation for all organizations. Rather, smart companies should invest in fixing and adapting their core technology before finding the best way to monetize innovation. Companies that get this right may indeed win big in the GenAI gold rush.
