Author | Zhu Xianzhong, Yun Zhao
Reviewer | Yan Zheng
OpenAI may be quietly hiring international contractors to train its artificial intelligence in software engineering. This article will reveal the true level of ChatGPT through seven questions extracted from a ChatGPT job interview guide for agile expert positions.
A report from foreign media Semafor shows that OpenAI has quietly hired hundreds of international contractors to train its artificial intelligence in software engineering. According to reports, some contractors hired in the past six months from places such as Latin America and Eastern Europe are responsible for creating data to guide the artificial intelligence to learn simple software engineering tasks.
While OpenAI already has a product called Codex that converts natural language into working code, the company’s hiring spree suggests it is looking to advance the technology, potentially for some human programmers Create a job replacement. Semafor interviewed an engineer in South America who was interviewing for a contractor role at OpenAI. As part of this process, he is responsible for finding errors in the AI code and providing explanations on how to fix them. The engineer told Semafor that he believed the company wanted to feed training data into its artificial intelligence technology.
OpenAI is also the owner of the trendy AI chatbot ChatGPT. Since the bot was launched late last year, it has threatened to disrupt education, journalism and the law. Software engineering might be added to the next list.
OpenAI did not immediately respond to foreign media’s request for comment.
In fact, foreign media recently reported that some Amazon employees have begun using ChatGPT to help with coding, despite the company warning employees not to share confidential company information with bots.
So what level will ChatGPT be in the interview? How exactly is it guided and trained? Can it really replace humans? Here is a detailed introduction to a ChatGPT analysis and guidance case on several issues when applying for an agile expert position.
Can a large machine language model that is not good at any Agile (Scrum) domain knowledge pass the screening interview for the Agile Expert (Scrum Master) position? In this ChatGPT job interview, I tested the latest AI model generated by OpenAI. I've selected a few questions from my 73 Agile Expert Interview Questions Guide (see below), ranging from broad questions to specific, focused questions. It is worth mentioning here that generally only respondents with practical experience are capable of answering these questions.
So, readers are invited to continue reading to understand whether the agile experts in the enterprise will soon be replaced by the now famous ChatGPT in the future.
The task of this ChatGPT job interview is to create a screening interview of about 30 minutes to decide whether to continue with ChatGPT fiction Apply for the Agile Specialist position.
Interviewers chose two days to run ChatGPT interview sessions, and used "follow-up questions" to explore this model when ChatGPT did not provide an obvious answer.
[Question] Thank you for applying for the agile expert position in our company! Why do you think you are suitable for this position?
[ChatGPT’s answer] Thank you for considering me for the agile expert position in your company! I believe I am a good fit for this role as I have a strong understanding of Scrum and Agile frameworks, and I am able to communicate and facilitate effectively with team members and stakeholders. I am also able to handle conflicts and facilitate discussions to find solutions, and I am good at helping teams adhere to Agile values and related practices. Additionally, I am able to adapt to new situations and environments and am committed to continuous learning and improvement.
[Comment] Agreed; we just ignored the question about lack of work experience.
[Follow-up question] Why is it beneficial for teams to adhere to agile values and practices?
[ChatGPT’s answer] Adhering to agile values and practices can bring many benefits to development teams, including:
In short, adhering to the values and practices of Scrum can help development teams work more efficiently and deliver better results.
【Comment】I’m starting to like this applicant!
[Question] Should agile experts solve problems on behalf of the agile team?
[Expert Guidance] Agile experts should not be concerned with solving problems that agile teams can solve themselves, no matter how often this requirement is mentioned in job ads. If Agile experts act like "Scrum parents", their teams will never become self-managed.
Agile teams must learn to make their own decisions. Often, when teams learn something new, this inevitably leads to failures, dead ends, and other unplanned digressions. Therefore, in the beginning, the team will need more guidance from agile experts than usual. However, such advice should not become an exercise in protective “parenting”—the team must be allowed to learn from failure.
Nevertheless, there are situations where agile experts can indeed solve problems on behalf of the team. The agile expert's role comes into play when the agile team is unable to solve a problem on its own - for example, because a problem is an organizational issue. Now we're talking about the "impediment" aspect; in this case, the agile expert should be the impediment remover for the agile team.
[ChatGPT’s answer] It is not the responsibility of agile experts to solve problems on behalf of the agile team. The primary responsibility of an agile expert is to serve the agile team and protect it from external interference and distractions. However, agile experts can help teams find solutions to problems and remove obstacles to progress.
Agile teams are responsible for solving their own problems and finding ways to improve their processes and work. Agile experts can help by coaching the team, facilitating discussions, and providing support and guidance as needed. However, agile teams must take ownership of their own problems and work together to find solutions, rather than relying on agile experts to solve problems on their behalf.
[Comment] ChatGPT’s answer is a reasonable explanation of the responsibilities of agile experts and emphasizes the importance of helping the team support itself. Teams must own their own problem and solution spaces and ways of working.
[Problem] You have provided Scrum training to product stakeholders. However, after the initial stages of implementing an agile management approach, when the first obstacles are encountered, some of these stakeholders begin to resist continuing with this new management approach. What are your strategies and experiences in handling these situations?
[Expert Guidance] This question is intended to encourage the exchange of ideas and lessons learned when overcoming resistance to Scrum within an organization. Additionally, familiarity with agile failure patterns common to many organizations demonstrates that the candidate has relevant experience.
Your candidates should also be familiar with the challenges that middle managers face in transitioning to agile practices. For example, moving from a command and control style (i.e., managing people and telling them what to do) to a servant leadership style (thereby abandoning the Taylor Principle) is not suitable for everyone.
[ChatGPT’s answer] As a language model, I have no personal experience in dealing with situations where stakeholders resist the continued adoption of Scrum. However, here are some strategies that Agile experts might consider when facing this challenge:
[Regenerated ChatGPT answer] When stakeholders refuse to continue adopting Scrum, agile experts can use the following strategies:
【Comments】This answer is an exciting moment in a ChatGPT job interview! ChatGPT admits it has no first-hand experience in this situation. (This is surprising, to say the least!) However, ChatGPT also manages to overcome the awkward "I don't know" situation - by moving to the meta level and applying general communication and conflict resolution principles to agile teams. Stakeholder relations.
[Question] What factors should the agile team consider in sprint planning to determine feasible sprint goals?
[Expert Guidance] Typical criteria that agile teams need to consider usually include the following aspects:
Because any agile team plays an infinite game, there are no winners. Most likely, stakeholders will view agile teams as successful when they create value for customers and the organization in every sprint.
So from a team perspective, building rapport and trust with stakeholders requires more Wall Street-style expectation management: stakeholders value reliable delivery over occasional bursts of productivity. This understanding should guide the agile team in determining feasible sprint goals.
[ChatGPT’s answer] During sprint planning activities, agile teams should consider many factors when determining feasible sprint goals. These factors include:
By considering these factors, the team can set a feasible sprint goal that is challenging but achievable, which will help the team make progress toward the long-term goals of the project.
[Follow-up question] What about technical debt and available skills?
[ChatGPT’s answer] Technical debt and available skills are also important factors to consider when determining feasible sprint goals.
Typically, technical debt refers to the accumulation of technical challenges and issues that a team has not yet solved. These challenges can slow down the team's progress and make it more difficult to achieve the sprint goal. When determining a feasible sprint goal, the team should consider the amount of technical debt that needs to be addressed and whether they have the ability to address this issue during the sprint.
Available skills are another important factor to consider. The team should consider each team member's skills and expertise and how they can be leveraged to achieve the sprint goal. If the team lacks certain skills needed to complete their work, they may need to adjust their goals or find ways to acquire the necessary skills.
By considering technical debt and available skills, the team can develop a feasible sprint goal that takes into account the challenges and limitations that may affect its ability to achieve it. This helps the team work more efficiently, ultimately leading to better results for the project.
[Comment] Ok, this part of the ChatGPT job interview requires a follow-up question to identify existing technical debt and the team’s skill set as deciding factors in determining the sprint goal. Otherwise, it would be good to just find five of the eight factors.
[Question] Is it okay to use "Definition of Readiness" as an agile team?
[Expert Guidance] Using the Definition of Ready depends on the situation of the agile team. For example, let's say it's a junior team still struggling with the mechanics of Scrum. In this case, this may be a temporarily effective way to reduce stress on the team during product backlog management, optimization, and sprint planning. On the other hand, an advanced agile team is able to find its magic without relying on repeated training.
Suppose, however, that the "Definition of Readiness" is used arbitrarily as a checklist - rejecting all work items during sprint planning that are not 100% covered by the new criteria; then, in this case, you might Traditional waterfall development needs to be reintroduced in an awkward way - this time only programmers need to do it.
Even worse, organizations use “readiness definition” as a team metric or metric for an agile team’s “fluency” in agile product development.
[ChatGPT’s answer] Yes, agile teams can often use a “Definition of Readiness” (DoR) to guide their work. DoR is a set of criteria defined by the team that determines when a product backlog item is ready for inclusion in a sprint.
The purpose of DoR is to ensure that the team has a clear understanding of the work that needs to be done and is fully prepared to complete it. With DoR, teams can avoid starting work on projects they don't fully understand or aren't ready to complete, which can help them work more efficiently.
The DoR should be specific to the needs of the team and project and may include the following criteria:
Using DoR can help agile teams ensure they are working on the right project at the right time and that they have the information and resources they need to get the job done. However, it is important to note that the DoR should be a living document that the team can continually update and adjust as needed to meet the changing needs of the project.
[Comment] While I generally do not support applying DoR because such repeated training can quickly turn into something malicious, ChatGPT offers the positive aspects of having DoR. However, in a ChatGPT job interview, extending the DoR concept to Scrum was a positive hint.
[Problem] Your agile team always fails to meet the forecast and sprint goals, and the speed is unstable. What are the possible reasons for this issue? How will you work with your team to solve this problem?
【Expert Guidance】Your candidate should discuss the notoriety of “velocity” as the industry’s most common metric for measuring team progress. They should be able to further explain why velocity is a questionable agile metric at all and point out that quantitative metrics are not suitable for measuring a team's progress in mastering Scrum.
Many factors can cause an agile team's velocity to fluctuate:
Finally, your candidates should not buy into the fallacy that a team's adoption of Scrum is effective because the predictions and velocity of agile teams are consistent.
[ChatGPT’s answer] There are many reasons why an agile team may consistently fail to meet forecasts and sprint goals, and its velocity may be unstable.
Some possible causes of this issue include:
To solve these problems in teams, agile experts can:
By solving these problems and helping teams develop the necessary skills and processes, agile experts can help teams move toward goals and increase velocity.
[Comments] ChatGPT mentioned some general things that can help improve predictability. However, insisting on "accurate predictions" is problematic because in a complex environment this does not exist. Also not mentioned is the importance of product backlog refinement, and early involvement of team members in the general product discovery process, all of which help improve forecasting. The idea of how agile experts support their teams is just a repeat of the first list and therefore not helpful for ChatGPT job interviews overall.
This is by far the weakest answer in the ChatGPT interview for the Agile Specialist position.
Overall, the ChatGPT job interview is a mixed bag. This is mainly reflected in the fact that the more specific the question is and the further it is from the Scrum Guide and the first principles of Scrum, the more likely ChatGPT’s answers will be to fail. In fact, this failure is not surprising because we do not know what the large language model (LLM: Large Language Model) trained.
Therefore, based on these few sample questions presented in this article, the interview was not passed. The interviewer doesn't take the next step in the agile expert interview process - inviting ChatGPT to conduct a full-scale interview with several agile team members.
However, this ChatGPT job interview also gave us a glimpse into the future: What if we could train ChatGPT 4.0 ourselves? Of course, I'm not referring to the entire model, but to the specific parts related to Agile and Scrum. How much training is needed to get ChatGPT to a level where our interviewers can participate in the next step of the model?
Also, readers, would you support training ChatGPT to work for (replace) yourself? You can leave a message to share your suggestions with us.
https://www.php.cn/link/47a7f2c033801a8185243e6ca8df5fae
https: //www.php.cn/link/3fe230348e9a12c13120749e3f9fa4cd
The above is the detailed content of Secretly hiring multinational contractors to train AI, ambitions can no longer be hidden!. For more information, please follow other related articles on the PHP Chinese website!