#34: How can GenAI "plug" this breach?

As I started writing this edition, I was reminded of something I wrote in wake of an earlier aviation incident:

Luckily, like that time, both the aviation mishaps of 2024 avoided a significant loss of life, reinforcing what I had said that time around:

Plan for the Best, Be Prepared for the Worst!

Otherwise things could have very well ended like the earlier Max disasters:

It was 2019 and I was then a bit worried about the role AI could have played in the disaster due to the botched MCAS software:

Now we have introduced another variable in the mix- AI, but, seems like have not clearly established its role in the hierarchy or defined the protocols to ensure that the hierarchy is adhered to. Does the MCAS give warning when it is taking over controls? Does it need to seek permission from the pilot to take over controls? Does the Pilot have the ability to easily take back control in a split second if needed ? Can the Pilot totally bypass the system if needed?

AI capabilities have grown leaps and bounds since then and Generative AI (GAI) is the new turbocharged kid in town now. This begs the question:

Can Generative AI prevent the recent Alaska Airlines Boeing 737 Max 9 type of disasters?

The tragic Alaska Airlines Boeing 737 Max 9 incident underscores the relentless pursuit of aviation safety. While a single solution may not exist, Generative AI (GAI) emerges as a promising tool with the potential to significantly reduce the risk of such disasters. Here's how:

Proactive Risk Mitigation:

• Predictive Maintenance: GAI can analyze data from aircraft sensors, anticipating component failures and enabling preventative maintenance, potentially averting in-flight emergencies. (Point to ponder: Could GAI have indicated an issue with the door plug as a probable cause of the three earlier pressurization warning light incidents?)

• Flight Simulation and Training: Immersive, GAI-powered simulations expose pilots to rare or challenging scenarios, honing their decision-making and response skills in critical situations.
• Design Optimization: GAI can assess countless design variations, pinpointing potential flaws and suggesting improvements, leading to safer and more resilient aircraft.

Real-time Support and Efficiency:

• Real-time Decision Support: In-flight GAI systems can analyze sensor data and environmental factors, providing pilots with critical insights and recommendations in real-time, guiding them through challenging situations.
• Automated Safety Checks: GAI can automate routine safety checks, eliminating the risk of human error and ensuring all systems are functioning optimally before and during each flight.

Post-incident Learning and Improvement:

• Accident Investigation: GAI can analyze flight data recorders and evidence with unparalleled speed and accuracy, pinpointing the cause of accidents far faster, informing future safety regulations and design modifications.

Beyond Individual Capabilities:

• Enhanced Air Traffic Control: GAI can optimize air traffic management, dynamically adjusting flight paths and reducing the risk of mid-air collisions, a potential source of human error in traditional systems.

It's crucial to remember that GAI isn't a magic bullet. The complexity of aviation and unforeseen circumstances can still present challenges. However, by integrating GAI's capabilities with human judgment and oversight, we can take a significant leap towards a safer future for air travel.

But all said and done human obstinacy and stupidy can still trump the logic of a well-trained model as manifested by the sad story of all the cockpit voice data from the Alaska Airlines flight being lost since it was written-over in a 2 hour cycle (this in a day and age when data storage is so cheap and non-bulky).

US airlines and regulators have no excuse other than sheer resistance to change.

The cockpit voice recorder data on the Alaska Airlines Boeing 737 MAX 9 jet which lost a panel mid-flight on Friday was overwritten, U.S. authorities said, renewing attention on long-standing safety calls for longer in-flight recordings.

Open Source vs Proprietary AI Models

In a recent LinkedIn post AI guru Yann Le Cun opined :

Open source AI foundation models will wipe out closed and proprietary AI models for the same reason Wikipedia wiped out generalist commercial encyclopedia: crowd-sourced human contributions to open platforms can cater to a high diversity of interests, cultures, and languages. - Yann LeCun

Interesting. This statement posits that open source AI foundation models will supersede closed and proprietary AI models, drawing a parallel to how Wikipedia outpaced commercial encyclopedias. Let's examine its key assumptions and implications:

1. Effectiveness of Crowd-Sourcing: The assertion hinges on the effectiveness of crowd-sourcing in enhancing AI models. While crowd-sourcing can indeed contribute to a wide range of perspectives and data, the quality and reliability of these contributions can vary. Unlike Wikipedia, where content can be easily moderated, integrating diverse data into AI models requires sophisticated quality control and curation.
2. Open Source vs. Proprietary Models: The statement assumes that open source models inherently have an advantage over proprietary models. However, proprietary models can benefit from dedicated resources, controlled data quality, and focused research and development efforts. They may also have commercial incentives to innovate and optimize.
3. Diversity and Inclusivity: Open source platforms might be more adaptable to diverse interests, cultures, and languages. However, this inclusivity depends on the active participation of a diverse group of contributors, which is not guaranteed. Proprietary models could also achieve this through targeted data collection and partnerships.
4. Economic and Resource Considerations: The comparison with Wikipedia overlooks the economic and resource aspects. AI development, especially foundation models, require significant computational resources, expertise, and continuous updates. Open source models might face challenges in securing consistent funding and resources.
5. Trust and Security: In some applications, closed proprietary models might be preferred due to concerns over security, privacy, and reliability. Open source models, due to their accessibility, might pose greater risks in these areas.
6. Market Dynamics and User Preferences: The success of an AI model in the market also depends on user preferences, commercial strategies, and partnerships. Wikipedia's success over commercial encyclopedias was influenced by its free access and the evolving internet culture, factors that might not directly apply to AI models.
7. Technological Advancements: The pace of technological advancements in AI could also influence this dynamic. Breakthroughs in AI might come from either open source initiatives or proprietary research, and either could tip the balance.

In summary, while the statement presents a compelling perspective, it oversimplifies the complex dynamics between open source and proprietary AI models. Factors such as quality control, resource allocation, market dynamics, and technological advancements play crucial roles in determining the success and adoption of AI models.

AI-Enabled Memories: A New Era of Grieving- Creepy or Cool?

Charlotte Jee 's article delves into the emerging technology of creating digital replicas of deceased loved ones using AI and voice cloning.

Developed by companies like HereAfter AI , this technology creates virtual versions from extensive personal data. Her experience with her parents' digital clones highlights the technology's potential for comfort, yet it also brings ethical dilemmas and psychological impacts to the forefront.

The technology, while offering a sense of continued connection, raises questions about consent, the prolongation of grief, and the emotional consequences of interacting with virtual family members.

As this technology evolves, it challenges us to consider the balance between the need for closure and the responsible use of innovative tools in processing grief. The article invites contemplation on the readiness of society to embrace a future where parting with loved ones is forever transformed by digital advancements.

This report from the Mayfield Fund distills insights from many IT and Innovation leaders about their outlook for 2024. The key takeaways offer lots of food for thought. (Thanks Gamiel Gran Shelby Golan )

Great to see my quote make it into the survey report:

"???? ???????? ???? ?????????? ???????? ???????? ???????????????????? ‘???????? ???? ????’ ???? ???????????????????? ‘???????? ???? ?????? ????.’" - Deepak Seth

Signing Off

We celebrated Karel Capek's Birthday this week (Jan 9). He's the one who gave us the term "Robot" in his Czech play: R.U.R

Here's how this plays out in the AI led future of ours:

Robots, oh robots, with circuits so bright,

In the AI future, a common sight.

They'll crunch the numbers, they'll clean our space,

But in the soul department, they'll never ace.

Karel ?apek chuckled, 'Miss Glory, see,

They're smart as whips, but soul-free.

In a world where AI reigns supreme,

It's the human soul that's the ultimate dream.

So let them plan, let them compute,

In their tireless, soulless pursuit.

For in a future where AI's king,

It's our laughter and tears that truly sing!

Keep an eye on our upcoming editions for in-depth discussions on specific AI trends, expert insights, and answers to your most pressing AI questions!

Stay connected for more updates and insights in the dynamic world of AI.

p.s. - The newsletter includes smart prompt based LLM generated content.