Apple can Dominate the Next Decade of AI

Apple White Paper:

In the wake of Apple’s strategic collapse to develop internal AI and iterate Siri, I thought it pertinent to discuss how Apple can not only catch up, but dominate in ways that no one can compete. 

Chips: efficiency and vertical integration

Apple being able to develop chips in house has enormous potential to iterate on AI efficiency. Current AI hardware is not able to utilize the new CRAM innovation which researchers reported up to 1000× reduction in energy consumption for AI processing using CRAM. This implies a future where memory scaling no longer lags behind compute scaling, addressing one of the biggest structural inefficiencies in modern AI systems. If Apple pivots to offering hardware that rapidly adapts to new AI hardware methods to align with software imposed limitations, they can further leverage their compute team for profit in a high margin space and further increase efficiency leads. Having the memory located directly on-package allows faster, and more efficient ram with massive capacities possible, which allows for efficient customization of ram limitations for agentic workflows in different industries and at different price points. 

The Endgame: New Hardware, Verified Agents, Certifications

The end goal is to offer a seamless and magical AI experience, where everything just works. Apple designed hardware will run local AI agents that utilize Apple licensed software to perform industry specific professional tasks such as Legal, Financial, or Bureaucratic; anything that can be easily automated. Applying a “Red Hat” approach to software allows an alternative approach to the SaaS models prevalent that fail when privacy of the underlying data requires local hardware or being air-gapped from the internet. 

Even the option to own your own hardware has latency and other benefits for professionals. In a legal context, putting a black box in the middle of a legal workflow is a rather risky move, especially when the black box is not liable for its output, the professional is. Local AI run on models with licensed software puts the control back in the hands of the business who can optimize their own models beyond the industry normal to suit the tone of their own firm. 

Apple can strategically solve the tone and monotonous result problems with incumbent AI strategies utilizing SaaS based strategies like Harvey AI. If every law firm uses Harvey AI, and Harvey uses the same underlying GPT-4 model, then every law firm has the same "intelligence." 

A law firm’s competitive advantage is its unique intellectual property and methodology. A centralized SaaS model flattens this advantage. Apple’s strategic approach under this white paper allows firms to inject their own precedents and style guides into local models, preserving their unique competitive edge.

This approach will give Apple several lucrative B2B opportunities. Firstly, selling AI hardware based on the latest research will lead to an inevitable upgrade cycle based on Moore’s Law. As compute expands exponentially, demands always seem to increase in step; therefore, it can be inferred that as AI technology advances, professionals will have to upgrade to the latest models to stay competitive in their industries on a regular basis. This leads to predictable sales on a steady upgrade cadence aligned with industry trends. 

In addition to hardware and verified agentic programs developed by professionals to streamline industries running on general purpose AI models, Apple can sell certification for AI competency. Utilizing training videos, company exhibits, and or training seminars to various professional industries. 

These certifications would be valuable to ensure that an employee will be able to quickly utilize the software at a new company so long as it follows the same general alignment of Apple hardware and Apple verified agentic workflows. This method: locks in high margin professionals to upgrade cycles on specialty AI hardware, gives professionals absolute privacy over their data in a world without that option that is easy to roll out, gets rid of a black box workflow problem in information critical industries such as law, and further locks in businesses and employees to utilize as many aspects of your product and software line as possible to decrease downtime with churn to train new employees.

Different packaging, same product

In the world of the philosophy of consciousness, most formal theories rely on what's referred to as an “extra factor.” This extra factor is used to describe why humans have free will, while everything around us seems to be determined. There needs to be something to separate us from the rocks that we don't assign agency to.

One interesting modern interpretation of this problem is the idea from Daniel Dennett who theorized that if consciousness is defined as “there has be to something that it's like to be you,” then an explanation of the extra factor could be that we narrate ourselves into existence.

From an evolutionary perspective, this idea is genius because it enables long term thinking. If I didn't think I'm going to exist in the next second because I have no sense of self, I won't make plans that don't satisfy an immediate urge but accomplished a more abstract long term goal.

AI is such an exciting field because we get to dive into how intelligence works. If something works in a machine, we will likely be able to use it to augment our own capabilities. Intelligence is intelligence, regardless of the source so I have to suspect that if our rules of logic and math are sound, ai will inevitably end up similar to humans as it's human tasks they are being trained on and given human tasks to accomplish.

Source: Researchers Discover AI Language Models Are Mirroring the Human Brain’s Understanding of Speech - The Debrief https://share.google/SAdRN7TxSsG2Rk3tR

Watch for sensationalist journalism

When I look at a news feed, I'm bombarded by all the half informed takes that don't provide context to prove a point. Yes, openai is losing money. So is every generative ai company at the moment. But how long did it take Uber to believe profitable? Facebook? Air BnB? or even all the new age sports gambling apps.

At least generative ai companies are trying to provide real value, instead of promote bad behavior. While ai is a tool that can be used to produce brain rot or detect cancer, a gambling app will never produce a statistically positive result. Yes, there is a lot of hype because there is potential value, but the blitz scaling used to rocket tech to the moon strategy didn't go away for the next silicon valley miracle.

Source: X https://search.app/qd8RC, https://youtu.be/Wcv0600V5q4?si=UKgEyZh28gxE_Vak