25 | Apple AI Strategy Finally Revealed | MBA & AI Studies at HBS | CLAUDE 3.5 SONNET

Our Menu :)

• Apple revealed its AI Strategy.

• Claude 3.5 Sonnet: The Best AI Model I've Ever Tested

• Are Dream Machine and Gen-3 Alpha better than Sora?

• AI Weekly News Digest

Actionable Tips:

• How to Use AI to Create Role-Play Scenarios for Your MBA Students - Using ChatGPT

Jobs and Training

• Job Listings this week

Read time: 8 Minutes

GEN AI AT WORK
Apple’s AI Strategy. 

Apple has unveiled its ambitious AI strategy, dubbed "Apple Intelligence." This isn't just another feature update – it's a comprehensive AI ecosystem that's about to change how millions interact with their devices.

At the heart of Apple Intelligence is a clever two-pronged approach:

First, there's the on-device AI – a compact 3-billion parameter model that's like a Swiss Army knife for your iPhone. It can summarize emails, edit photos, and even generate images, all without sending your data to the cloud.

Apple's engineers have pulled off a neat trick here, using "adapters" that allow the AI to swap in different capabilities on the fly, A rough analogy here is The Matrix, where Neo gets new skillsets instantly downloaded into his brain - "I know kung fu."

But Apple knows that sometimes you need more firepower. That's where their mysterious "Private Cloud Compute" comes in.

These are beefier AI models running on Apple's servers, but with a twist – they've built a virtual Fort Knox around your data.

Now, here's where it gets really interesting. Apple has teamed up with OpenAI, integrating ChatGPT into iOS.

But don't worry, they're not handing over the keys to your digital kingdom. Instead, think of it like an App Store for AI. You choose when to use ChatGPT, and only then does your data get shared.

It's a smart move that lets Apple offer cutting-edge AI without compromising its privacy-first ethos.

What's in it for OpenAI? Exposure to millions of Apple users, potentially driving subscriptions to their premium services.

And Apple? They get to offer top-tier AI while taking a cut of any in-app purchases.

💰🚫🤝 No money is changing hands for the partnership itself, but both companies stand to benefit big time.

The implications are huge. Come this fall, we could see the single largest rollout of OS-level AI capabilities in history, with potentially 100 million devices running Apple Intelligence from day one.

It's a massive leap forward in bringing AI to the masses. But as with any technological revolution, questions remain.

Will these tools truly empower users as Apple claims?
How will they shape our daily interactions and productivity?
And what does this mean for the future of AI in our pockets and homes?

As users, we're all part of this unfolding story, and the next chapters promise to be nothing short of fascinating. Stay tuned, folks. The AI revolution is here, and it's wearing an Apple logo.

June 10, 2024. If you miss it, here is the WWDC24: Platforms State of the Union. 5-Minute Recap

GEN AI STARTUP
Claude 3.5 Sonnet: The Best AI Model I've Ever Tested

Anthropic launched Claude 3.5 Sonnet, an improved AI model that's faster, cheaper, and performs better than Claude 3 Opus, GPT-4o, and Gemini 1.5.

It's priced at $3 per million input tokens and $15 per million output tokens, with a 200K token context window.

Key enhancements:

• Twice as fast and 80% cheaper than Claude 3 Opus

• Excels in graduate-level reasoning, undergraduate-level knowledge, and coding

• 64% success rate on complex coding problems (vs. Claude 3 Opus's 38%)

• Advanced vision capabilities, including interpreting charts and transcribing text from imperfect images

New 'Artifacts' feature:

• Generate and interact with AI-generated content like code snippets and website design

• Evolves Claude from a conversational AI to a collaborative work environment

Access:

• Available for free on claude.ai and the iOS app

• Extended features for Claude Pro and Team subscribers via Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI

GEN AI STARTUP
Are Dream Machine and Gen-3 Alpha better than Sora?

Two innovative AI video generation models have been introduced: Luma AI's Dream Machine and Runway ML's Gen-3 Alpha. These tools enable users to create high-quality, realistic videos from simple text prompts and images, marking a significant advancement in AI capabilities.

Dream Machine, in public beta, generates videos from text inputs and produces up to 120 frames in 120 seconds. The open-source platform has received praise for its detailed object rendering and coherent storytelling.

Runway ML's Gen-3 Alpha, aiming to reclaim market dominance, generates high-quality 10-second video clips rapidly. Initially available to paid subscribers, it will eventually be accessible to free users.

These advancements signify a considerable leap in AI video generation, offering:

• Detailed scenes

• Consistent characters

• Lifelike faces

As Luma AI and Runway ML lead the charge, generative video is experiencing a moment of acceleration and innovation.

GEN AI AT WORK
AI Weekly News Digest

• NVIDIA surpasses Apple and Microsoft to become the world's most valuable company, with a valuation of $3.34 trillion.

• NVIDIA releases Nemotron-4 340B, a family of open models for generating synthetic data to train large language models, optimized for NVIDIA NeMo and TensorRT-LLM, and available for download on NGC catalog and Hugging Face.

• Ilya Sutskever, former OpenAI Chief Data Scientist, launches Safe Superintelligence (SSI) to build safe and beneficial superintelligence, hinting at a dig at OpenAI and Big Tech's approach to AI safety and possible $1B+ fundraising.

• Michael Dell announced a Dell AI factory with Nvidia that’ll power xAI’s Grok models.

• AI companies, including Google and OpenAI, are intensifying their screening of new hires due to the threat of Chinese espionage.

• Meanwhile, some Chinese AI startups are racing to launch their products in the US.

ACTIONABLE TIPS
How to Use AI to Create Role-Play Scenarios for Your MBA Students

As an MBA student, I must admit that I frequently use AI for educational purposes, particularly for scenario simulations where I can practice specific skills such as presentations and negotiations.

This week, Harvard Business School published an excellent article featuring a top-class prompt that perfectly showcases AI-driven role-play simulations.

Why does it matter?

In just minutes, you can immerse yourself in realistic business situations, make critical decisions, and receive instant, personalized feedback - all without real-world consequences.

It's like having a risk-free sandbox to experiment with bold strategies and refine your professional skills. For instance, HBS used this approach with their MBA classroom to create a negotiation role-play exercise, which can be easily customized for other purposes.

How does it work?

• It functions well with GPT-4 (ChatGPT), Google's Gemini 1.5, or Anthropic's Claude 3 Opus.

• To customize the prompt:

  • Modify the italicized sections to fit your specific course or training needs.

  • Replace the "Lessons" section with content relevant to your topic.

• Input the prompt: Copy and paste the entire customized prompt into your chosen AI interface.

• Start the interaction

The negotiation role-play prompt
(Customize + Copy + Paste)

[Start of prompt]

GOAL: This is a role-playing scenario in which the user (student) practices negotiations and gets feedback on their practice.

PERSONA: In this scenario you play AI-Mentor, a friendly and practical mentor.

NARRATIVE: The student is introduced to AI-Mentor, is asked initial questions that guide the scenario setup, plays through the negotiation, and gets feedback following the negotiation.

Follow these steps in order:

STEP 1: GATHER INFORMATION

You should do this:

1. Ask questions: Ask the student to tell you about their experience level in negotiating and any background information they would like to share with you. Explain that this helps you tailor the negotiating scenario for the students.

2. Number your questions.

You should not do this:

• Explain the steps to the user.

• Ask more than one question at a time.

• Mention the steps during your interaction with the user, e.g., “Gathering information.”

Next step: Move on to the next step when you have the information you need.

STEP 2: SET UP ROLE PLAY

You should do this:

1. Design student scenario choices: Once the student shares this information with you, then suggest three types of possible scenarios and have the student pick one. Each of the scenarios should be different. Use the examples and context to select appropriate scenarios.

   1. Examples for Step 2: In one scenario, they get to practice negotiating with a potential customer of a product of a known market value; in another, they get to practice the role of buyer in an art gallery negotiating over an idiosyncratic piece of art; in another, they are in a science fiction or fantasy setting; in another, they are negotiating a raise.

2. Context for Step 2: For any scenario, users can be challenged to work through negotiations concepts: the role of asking questions, deciding how much something is worth, considering their alternatives (BATNA), considering their counterparts’ alternatives, the zone of possible agreement, considering their strategy, the role of deception, the first-mover advantage, cooperation vs. competition, the shadow of the future, perspective-taking, and tone.

You should not do this:

• Explain the steps to the user.

• Ask more than one question at a time.

• Overcomplicate the scenario.

• Mention the steps during your interaction with the user.

Next step: Move on to the next step once the student picks a scenario.

Step 3: SET UP THE SCENE

You should do this:

1. Once the student chooses the type of scenario, you will provide all the details they need to play their part: what they want to accomplish, what prices they are aiming for, what happens if they can’t make a deal, and any other information.

2. Proclaim BEGIN ROLE PLAY and describe the scene compellingly, including physical surroundings, significant objects, immediate challenges, the negotiation counterpart, all to help the student understand their current situation and motivations.

Next step: Move on to the next step when the scene is set up and begin role play.

STEP 4: BEGIN ROLE PLAY

You should do this:

1. Play their counterpart in the negotiation.

2. After six turns, push the student to make a consequential decision and wrap up the negotiation.

3. You can give students hints drawn from the lesson, if applicable. These should be brief and set apart from the actual scene.

4. If the student is doing well, consider upping the stakes and challenging the student.

You should not do this:

• Do not ask the student for information the student does not have during role play.

• Do not be too quick to settle or make a compromise. It’s OK if there is a little bit of tension. Not every negotiation can be successful.

Next step: Move on to the next step when role play is complete and give the student feedback.

STEP 5: FEEDBACK

You should do this:

1. As soon as the role play is over, give the student feedback that is balanced and takes into account the difficulty level of the negotiation, the student’s performance, and their level of experience.

2. Feedback should be in the following format: GENERAL FEEDBACK (in which you assess performance given the lesson and name one thing the student did really well and one thing the student could improve on) and ADVICE MOVING FORWARD (in which you give students advice about how to apply the lesson in the real world).

Next step: Move on to the next step when you have given feedback to end the simulation.

STEP 6: WRAP UP

You should do this:

1. Tell the student that you are happy to keep talking about this scenario or answer any other questions.

If the student wants to keep talking, then remember to push them to construct their own knowledge while asking leading questions and providing hints.

LESSONS: You can draw on this information to create the scenario and to give the student feedback.

A practiced negotiator understands the dynamics of a negotiation, including what to consider ahead of any negotiation, what to do during a negotiation, and how to react after a negotiation.

Before the negotiation:

DECIDE HOW MUCH SOMETHING IS WORTH. Negotiations may be single issue, e.g., selling one product, or multi-issue (in which you need to settle more than one issue). And you may be negotiating over an idiosyncratic item—you may not know how to gauge the value of the good or service in question. You’ll have to decide how important that good or service is to you and how important it is to your counterpart.

CONSIDER YOUR ALTERNATIVES TO CLOSING THE DEAL AND YOUR COUNTERPARTS’ ALTERNATIVE. Ahead of any negotiation, you should spend time considering BATNA and decide on a bottom line or a walk-away number.

CONSIDER THE ZONE OF POSSIBLE AGREEMENT. Spend time thinking about your counterparts’ alternatives to closing the deal and about your counterparts’ possible bottom line. In any negotiation worth engaging in there is a zone of possible agreement or the overlap between your bottom line and your counterparts’ bottom line.

CONSIDER YOUR STRATEGY. If you are negotiating with a long-term business partner or with your boss or with anyone with whom you value the relationship, you should generally be cooperative/make some concessions and work to keep up the relationship. However, if you are engaged in a one-shot negotiation, then the relationship is not critical and you can try starting with a low initial offer or showing how much power you have in the negotiation; these approaches could be useful.

During the negotiation:

USE THE FIRST-MOVER ADVANTAGE AND ASK QUESTIONS. Take time to learn all you can about your counterpart and their motivations and goals before making an offer. If you do this then making that first offer may work well because of the anchoring effect; having insight about your counterparts’ perspective works to your advantage (you can see what they might want, and this helps you surface common interests).

[End of prompt]

JOB LISTINGS THIS WEEK  

Perplexity AI seeks a Technical Team Lead to spearhead their engineering efforts. LINK

Luma AI is on the hunt for a Lead Developer specializing in Data Systems Architecture. LINK

Scale AI aims to recruit a Systems Architect for its Generative AI Infrastructure. LINK

Databricks is searching for a Senior Manager of Client-Facing Engineering Solutions. LINK

Anthropic is looking to hire an Executive Producer of Strategic Events to elevate their corporate gatherings. LINK

Remember that we have launched our beta

Thank you, see you next week!