This lecture provides in-depth advice on career development in the AI field. Professor Andrew Ng emphasizes that we are in a golden age of AI advancement, explaining that the development of AI building blocks and AI coding tools has dramatically accelerated software development. He particularly highlights the "product management bottleneck," stressing the growing importance of engineers who can also participate in product planning.

Guest speaker Laurence Moroney then diagnoses the current AI job market and presents three key pillars for success in AI: deep understanding, business focus, and a bias towards delivery. He also explains the importance of "vibe coding" and technical debt management, critiques the AI industry's "hype cycle," and recommends a measured approach. Notably, he predicts a bifurcation between large-scale and small-scale AI models in the future, with Small AI creating new opportunities.

1. AI Career Development: Now Is the Best Opportunity!

Professor Andrew Ng emphasized that now is the best time to develop and build a career in AI. Although questions had been raised just a few months earlier about whether AI progress was slowing, he offered an optimistic outlook.

"The complexity of tasks AI can do is doubling every 7 months. For AI coding, this cycle is much shorter at 70 days."

This is measured by the complexity of tasks AI can handle and the time it takes humans to perform those tasks, indicating that the pace of AI advancement is extremely rapid.

1.1. A More Powerful and Faster Development Environment

Professor Andrew Ng identified two key drivers of AI advancement:

1. More powerful AI building blocks: Advances in LLMs, RAG (Retrieval-Augmented Generation), voice AI, deep learning, and other AI building blocks have made it possible to create software of previously unimaginable power.
2. Dramatic advances in AI coding tools: Thanks to AI coding tools, the speed of writing software has become unprecedentedly fast. He personally cited tools like Claude Code, OpenAI's Codex (post-GPT-5), and the recently released Gemini 3, noting that these tools advance rapidly every 3--6 months. He stressed that using the latest versions of these tools is critical for maintaining productivity.

"Being half a generation behind on these tools honestly means your productivity drops quite a bit."

1.2. The Product Management (PM) Bottleneck and Engineers' New Role

With AI coding advances making software development much easier and faster, Professor Ng pointed to a new bottleneck: the Product Management Bottleneck.

"As going from a clearly written software spec to code becomes easier and easier, the bottleneck is increasingly shifting to deciding what to build, or writing a clear spec for what you actually want to build."

The engineer-to-product manager (E:PM) ratio, previously around 4:1 or 7:1, is now shrinking to 2:1 or even 1:1. This signifies the growing importance of the product manager role.

Professor Ng emphasized the importance of engineers developing the ability to define products. He shared his experience of regretting asking engineers to take on more product-related work early in his career, yet noted that the fastest-moving people today are engineers who can communicate with users, gather feedback, and decide what to build based on deep empathy.

1.3. The Importance of the People You Work With

For career development, Professor Ng places great importance on the influence of the people around you.

"One of the strongest predictors of speed of learning and level of success is the people you surround yourself with."

He highlighted the value of Stanford's rich connective tissue, noting that many working in cutting-edge AI labs were former students of Stanford faculty. Through such networks, one can quickly learn about information and technology trends not yet widely known.

He also advised choosing workplaces based on the quality of teammates rather than a company's "hot brand." Through the example of a Stanford graduate who joined a famous company expecting to work on AI projects but was assigned to a backend Java payment processing team and eventually left, he illustrated the importance of team and colleagues.

"Rather than working at the company with the hottest brand, I think sometimes finding a team with a really good group of hardworking, knowledgeable, smart people trying to do good things with AI -- even if the company logo isn't as famous -- is the path to learning faster and advancing your career better."

1.4. Taking Bold Action with Responsibility

Professor Ng encouraged bold experimentation with responsibility now that powerful software can be built faster than ever.

"You have to be responsible about not building software that harms others. At the same time, there is so much that each of you can build."

He noted that there are many ideas in the world but a shortage of people with the technical skills to build them, and advised not hesitating to try various things since the cost of failure is low (just losing a weekend and learning).

Finally, while acknowledging some might consider it "politically incorrect," he encouraged working hard. While respecting those whose personal circumstances make this difficult, he stated that for those fortunate enough to be in a position to work hard, enormous opportunities exist right now.

"If you're in a position in life where you can work really hard, there is so much you can do right now." "If you're as excited as I am about coding in the evenings and weekends and building things and getting user feedback, if you lean into doing those things, your odds of success will increase."

2. AI Job Market Realities and Success Strategies: Laurence Moroney's Advice

Following Professor Ng's introduction, bestselling AI author and award-winning researcher Laurence Moroney continued with AI career advice.

2.1. The Corporate Perspective: Looking for 'People We Want to Work With'

Laurence Moroney added to Professor Ng's advice about carefully choosing who to work with, emphasizing that companies also select people they want to work with. He illustrated this through the story of a young job seeker he mentored over 18 months.

This young man, with an excellent educational background and coding skills, had been laid off from a medical software company the previous April, compounded by a breakup and the death of his dog. He applied to over 300 companies and passed coding tests perfectly at top firms like Meta, Microsoft, and Jeff Bezos's company (Amazon), yet was rejected every time.

Through mock interviews, Moroney identified the problem. The young man had interpreted the hiring guide's advice to "stand your ground and be confident" as behaving "really firm and strong." This came across as an adversarial attitude to interviewers, leaving the impression that even with excellent skills, "we wouldn't want him anywhere near our team."

"The advice he received to 'stand your ground' ultimately made him adversarial in the interview environment."

After Moroney's coaching to adjust his attitude, the young man was hired at a company that values teamwork, at twice his previous salary. This case clearly demonstrates that attitude and soft skills for collaboration are just as important as outstanding technical skills.

"If you get advice from tech interview coaching to 'stand your ground and be confident,' that's fine, but while doing so, don't be a jerk."

2.2. AI Job Market Reality: Opportunities Amid Challenges

Moroney objectively analyzed the current AI job market reality:

• Junior hiring slowdown: Entry-level graduate hiring has noticeably decreased.
• Big tech layoffs: Headlines are dominated by large-scale layoffs at Google, Amazon, Microsoft, and others.
• Rising entry barriers: There's a feeling of insufficient entry-level positions. (But this is just a feeling; looking more closely reveals a different picture.)
• Intense competition: Competition is very fierce.

Despite this, he stressed that there's no need to worry.

"No. If you can approach it the right way, especially if you understand how fast the AI landscape is changing, people with the right mindset will thrive."

2.3. AI Industry Changes: From the Era of Over-Hiring to Reality

The AI industry has undergone dramatic changes:

• 2021--2022: Due to the global pandemic's industry slowdown and shift to remote work, companies focused on areas directly contributing to revenue, reducing hiring.
• 2022--2023: After the pandemic, AI exploded and every industry shifted to AI-first, triggering massive hiring. Over-hiring occurred, with underqualified individuals filling senior positions and people being hired simply for having AI experience.
• 2024--2025: The period of the "Great Wakeup" -- companies recognized the problems of over-hiring and underqualified staff, becoming more cautious in hiring.

Moroney noted that understanding this situation and approaching it strategically reveals abundant opportunities.

2.4. Three Pillars for Success

Moroney presented three key pillars for succeeding in the AI business world:

1. Understanding in Depth

• Academic understanding: Possessing deep academic knowledge of ML, specific model architectures, paper comprehension, etc., and the ability to apply it in practice.
• Trend insight: The ability to identify where the signal-to-noise ratio is high within specific AI trends -- finding the truly important signals amid vast information.

2. Business Focus

• Measurable effort: Rather than the vague concept of "working hard," it's important to measure the value of effort through output.

  "Hard work isn't about how many hours you put in, but how much value you create."

• Output for your desired role: Creating and demonstrating output aligned with the role you want. Moroney shared his experience of developing a Java cloud application and putting it on his resume when applying for a Google engineer role, which allowed him to lead the interview when asked about his code.

  "Don't work for the job you have; create output for the job you want."

3. Bias Towards Delivery

• Ideas are cheap. Execution is everything: Even the best ideas are meaningless without execution. The ability to produce concrete results is what matters.

  "Ideas are cheap. Execution is everything."

2.5. AI in Practice: The Shift to 'Usefulness' and 'Productivity'

In the past, making "cool things" was important, but now making "useful things" is what matters, Moroney emphasized. Everything is oriented toward production, and companies are optimizing for profitability.

"Up until now, if you could do cool stuff, great. If you could build an image classifier, it was a golden era. Hundreds of thousands of dollars in salary and huge stock options. Unfortunately, that's no longer the case."

2.6. Four Realities: Business, Risk, Responsibility, and Learning from Mistakes

He presented four realities facing AI professionals:

1. Business focus is non-negotiable: Unlike the era when Silicon Valley companies emphasized individual employee value and "bringing their entire self to work," focus on business has become the top priority. Companies have returned to prioritizing profitability and business outcomes after "excessive activism" negatively affected business.
2. Risk mitigation is part of the job: The ability to understand and mitigate risks that may arise during AI-driven business process transformation is a critical skill. Demonstrating this risk management mindset in interviews is important.
3. Responsibility is evolving: The definition of responsible AI is shifting from vague social issues to specific problems that directly impact business and don't damage corporate reputation. Moroney illustrated through Google Gemini's image generation error (failing to generate white female images) how poorly implemented safety filters can damage corporate reputation.
4. Learning from mistakes: Mistakes are inevitable in AI development, and continuously learning and improving from mistakes is essential. Being forgiving when colleagues make mistakes and working together to solve problems is also important.

2.7. 'Vibe Coding' and Technical Debt Management

Moroney acknowledged that while Professor Ng doesn't love the term "vibe coding" (AI generating code), he would continue using it, dispelling the misconception that it makes engineers less useful.

"The more skilled the engineer, the better they use this type of 'prompt coding.'"

He explained the concept of "technical debt," noting that all software development incurs debt, and managing it wisely is key. Using the analogy of "good debt" from home purchases versus "bad debt" from high-interest credit card purchases, he advised creating "good technical debt" in software development by generating business value, writing understandable code, and setting clear goals.

Three key elements for good technical debt:

1. Clear goal setting and achievement: "Knowing what you need to build, and not just firing up ChatGPT and spinning code."
2. Delivering business value: "How does this help the business? Is this actually driving something?"
3. Human understanding: "The worst technical debt is delivering code that nobody can understand." -- Clear documentation, algorithms, variable names that enable others to understand the code.

Examples of bad technical debt:

• Solutions looking for problems: Using tools without thinking about the problem.
• Spaghetti code: Poorly structured code, especially from repeated prompting.
• Authority override: When leadership generates code from personal curiosity that the team then has to maintain.

2.8. Navigating the AI Hype Cycle

Moroney warned about the hype cycle pervasive in the AI industry. Social media values engagement over accuracy, resulting in an abundance of misinformation about AI.

"The currency of social media is engagement. Accuracy is not the currency of social media."

The ability to filter signal from noise within the hype is critically important and will make you a uniquely valuable professional.

He shared an experience from last year when a European company requested "agent" implementation. Starting with the fundamental question "why do you want to build an agent?", he ultimately discovered the company's true goal was "increasing sales team efficiency." He explained the four stages of agentic AI (understand intent -> plan -> execute -> reflect on results), emphasizing that deep understanding of business requirements and careful problem-solving are the keys to successful AI projects.

"If a company comes and says 'implement an agent for us,' the right first question to ask is: 'Why?'"

Strategies for becoming a trusted advisor amid the hype:

• Objectively evaluate trends: Distinguish between what's fashionable and what's a real opportunity.
• Deep understanding of fundamentals: Develop the ability to explain technical reality to non-experts.
• Make it as mundane as possible: Practice explaining new technology in the most ordinary, understandable way.
• Continuously track developments: Find important signals even amid social media's "cess pits."

2.9. The AI Bubble and Future Outlook: Big AI vs. Small AI

Moroney predicted that an AI "bubble" is likely coming. He warned of excessive hype, declining VC investment, unrealistic valuations, and proliferating me-too products -- similar to the dot-com bubble. However, just as Amazon and Google survived and thrived after the dot-com bust, companies and professionals who stay true to fundamentals, build real solutions, and understand the business side will survive and prosper through the AI bubble.

He predicted that within 5 years, the AI industry will bifurcate into "Big AI" and "Small AI".

• Big AI: The direction of current large language models (Gemini, Claude, OpenAI) growing larger and more advanced toward AGI.
• Small AI: Based on open-source models (open weights, self-hostable), with individuals and companies self-hosting and fine-tuning for specific tasks. He explained this will present major opportunities in fields where privacy and intellectual property protection matter (law firms, medical institutions, film studios, etc.).

  "Today's 7B model is as smart as yesterday's 50B model. Next year's 7B model will be as smart as last year's 300B model."

Working at ARM, he believes a world of "AI everywhere all at once" is coming, with technologies like SVE (Scalable Matrix Extensions) in the mobile space enabling AI workloads to run on CPUs, creating new scenarios for implementing AI on low-power devices beyond the traditional CPU/GPU computing paradigm.

"Breaking the habit that GPUs must be the ones doing AI is part of the trend the world is heading towards. Apple is probably one of the leaders in that space."

2.10. 'Artificial Understanding' and Developing 'Superpowers' Through AI

Finally, Moroney discussed a hidden aspect of AI: "artificial understanding." When models understand situations on our behalf and create new things based on that understanding, we can develop "superpowers" far more effective than before.

He showed an example where he prompted AI to create a video of his son scoring a goal from a photo of an ice hockey shot, but the AI added spectators and gave him two sticks -- incorrect results. This was because the agentic workflow of understanding intent, planning, executing, and reflecting was not followed.

In contrast, he demonstrated a film production demo using virtual actors developed with a startup, successfully expressing virtual actors' emotions and generating footage matching the story context through an agentic workflow. This suggests that when AI deeply understands the user's intent and considers tool characteristics to plan accordingly, it can produce far more accurate and useful results.

Conclusion

This lecture provides deep insights and practical career advice about the present and future of the AI field. Professor Andrew Ng emphasized that we are in a golden age of AI, arguing that AI building blocks and coding tools have made software development more powerful and faster. He particularly highlighted the product management bottleneck, the importance of engineers developing product planning capabilities, and the value of working with great colleagues.

Laurence Moroney diagnosed the AI job market reality, advising that careers should be built on three pillars beyond mere technical skill: deep understanding, business focus, and a bias towards delivery. He emphasized the importance of finding signal amid the hype, wisely managing technical debt, and acquiring the skills needed in a future AI industry bifurcated into "Big AI" and "Small AI." Ultimately, AI will grant humans new superpowers through "artificial understanding" beyond being a mere tool, and those who understand and strategically respond to this current of change will lead the future AI era.