Tag: AI

Part 2: From Arches to Pipelines

The genius of Roman engineering wasn’t just in their monuments—it was in their patterns.

  • Arches distributed heavy loads with elegance.
  • Domes enclosed vast spaces without collapsing.
  • Concrete gave them strength, flexibility, and the ability to scale construction.

These patterns were reusable, adaptable, and reliable—allowing Rome to expand from one city into an empire.

In the digital world, AWS generative AI applies the same principle of reusable patterns:

  • SageMaker pipelines are today’s arches—distributing workflows, balancing complexity, and channeling resources efficiently.
  • Bedrock APIs are modern domes—enclosing sophisticated models in simple, accessible interfaces.
  • Inferentia and Trainium chips are the new concrete—providing a durable foundation of performance and efficiency.

Both Rome and AWS solved the same problem: how do you build something that scales reliably without reinventing from scratch every time?

Great design is timeless—whether in stone or in code.

From Arches to Algorithms: Foundations Across Time

When we think of Roman architecture, what comes to mind? Colosseums, aqueducts, and basilicas—structures that stood the test of time. The Romans weren’t just building for beauty. They engineered for symmetry, durability, and public utility. Their aqueducts carried water across miles with remarkable precision, and their basilicas and forums became centers of civic life and governance.

Now, fast forward nearly 2,000 years. Today’s architects of generative AI face a very different medium—code and cloud instead of stone and marble—but the design questions aren’t so different.

In the world of AWS generative AI, the foundations are about scalability and modularity. Instead of concrete and arches, we build with services like:

  • Amazon SageMaker for streamlined training and deployment, bringing together widely adopted AWS machine learning (ML) and analytics capabilities, the next generation of Amazon SageMaker delivers an integrated experience for analytics and AI with unified access to all your data. Collaborate and build faster from a unified studio using familiar AWS tools for model development in SageMaker AI (including HyperPod, JumpStart, and MLOps), generative AI, data processing, and SQL analytics, accelerated by Amazon Q Developer, the most capable generative AI assistant for software development. Access all your data whether it’s stored in data lakes, data warehouses, or third-party or federated data sources, with governance built in to meet enterprise security needs.

  • Amazon Bedrock for direct access to generative AI models via APIs. Amazon Bedrock is a comprehensive, secure, and flexible service for building generative AI applications and agents. Amazon Bedrock connects you to leading foundation models (FMs), services to deploy and operate agents, and tools for fine-tuning, safeguarding, and optimizing models along with knowledge bases to connect applications to your latest data so that you have everything you need to quickly move from experimentation to real-world deployment.

  • AWS Inferentia chips to deliver cost-efficient performance at scale. AWS Inferentia chips are designed by AWS to deliver high performance at the lowest cost in Amazon EC2 for your deep learning (DL) and generative AI inference applications. 

Just as Roman engineers thought about structures that would last for centuries, AWS engineers design digital systems that can scale globally, adapt instantly, and endure change.

The underlying truth is timeless: whether in stone or in cloud, strong foundations determine what endures. Rome’s enduring arches echo in today’s scalable pipelines. Both ask the same question: what can we build today that will still matter tomorrow?

Weekly Roundup Of Tech News – 05/23/2021

Healthcare: Telephones not computers played key role in pandemic TeleHealth 

What: According to Axios KFF survey reported that more than half of Medicare beneficiaries utilized telephone for their Telehealth visits.

How:  More than 56% of beneficiaries used telephone for the Telehealth. It was very high among hispanics (61%), rural (65%) compared to only 28% of people using video for telehealth.

Why it matters: Telemedicine conjures up video visits from the physician but this statistic provides an insight into the adoption by the end consumer. There could be challenges in the availability of broadband to the rural and minority communities. Until these challenges are addressed adoption of Telehealth will continue to be a challenge.

Artificial Intelligence: RAI’s certification to prevent AI turning into HALs

What: Responsible Artificial Intelligence Institute (RAI), a non-profit hopes to offer a more standardized means of certifying AI solutions.

How:  RAI has built a concrete framework of Build, Accredit, Audit and Certify process that has dimensions in Accountability, Bias and Fairness, Data Quality, Explainability and Interoperability and Robustness for Certifying AI solutions.

Why it matters: We have seen how AI’s can go rogue through in the fictional Space Odyssey’s HAL computer which eliminates the entire crew. More recently, Microsoft’s Tay debacle, Facebook’s algorithm spreading online hate and the Clearview’s surveillance systems’ facial recognition software caused public outrage due to their power and the opaqueness of algorithms’ logic creating fear about AI itself. By certifying the AI systems similar to LEED, it gives transparency and more adoption.

Worldwide Web: Linkrot and its impact on the web 

What: Research has shown that many important links in the web get lost to time. For e.g., quarter of The New York Times’ articles are now rotten, leading to completely inaccessible pages according to team of researchers from Harvard Law School. The following graph shows reverse view of link rot over time.

How: When an old article gets archived, the new location is not published. For example, let’s say an article was published in 1998 with a hard code the link and has been archived. The original link wouldn’t be active and someone else can publish a completely opposing view of the original content thereby affecting the integrity of the content. The study by Harvard Law School found that in 550,000 articles, which contained 2.2 million links to external websites  in New York Times, 72% of them were “deep” or pointing to a specific page rather than a general website. 6% in 2018 vs 72% links from 1998 were dead.

Why it matters: Imagine a situation where the original video or content succumb to linkrot and in its place something else is published that could create confusion and panic. One solution is by Wikipedia where it asks for page’s archive on sites like Wayback machine. Another solution by Perma.cc project attempts to fix the issue of link rot in legal citations and academic journals by providing archived versions of the page along with original source. There are many other areas that require this capability and certainly something for a startup to think about. Any takers?

Weekly Roundup Of Tech News – 5/9/2021

  1. Software Development: US Supreme Court Rules on Key Software Development Practice
    • What: Supreme Court ruled in favor of Google about its usage of Java SE compatible programming interface for Android Development as “Fair Use” in a case filed by Oracle.
    • How: Even though Google used about four-tenths of a percentage of Java Code, and that such code was further incorporated into a totally different product was transformative enough use of the code at issue to qualify its fair use of that code.
    • Why it matters: What this means for the developers is that they can continue utilizing the open source APIs with the understanding that implementations matter more than the definition. As such a big win for the development community.
  2. Artificial Intelligence: USPS turns to AI to boost Package Processing
    • What: USPS handles roughly 129 billion pieces of mail and 7.3 billion packages last year. Tracking these have been difficult. A federal data scientist proposed to deploy the edge AI servers at the postal servicing processing centers system in an effort to gain and share more data points.
    • How: NVDIA working with USPS created Edge Computing Infrastructure Program or ECIP, a distributed edge AI system now running at USPS locations, via NVDIA EGX Platform.
    • Why it matters: The system used to take 8 to 10 people several days to track down items now takes one or two hours with the ECIP AI Program. This enables USPS to track down any item in transit to better manage the deliveries.
  3. Cryptocurrency:  Digital Dollar Project to launch currency pilots
    • What: U.S Nonprofit Digital Dollar Project said it will launch five pilot programs over next 12 months to test use cases for US Central Bank Digital Currency (CBDC)
    • How: Private-sector pilots are funded by Accenture Plc and involve financial firms, retailers and NGOs to generate data to help US Policy makers develop digital dollar through central banks.
    • Why it matters: The data derived can pave way for creating US CBDC that will assert its place as a digital currency and will help larger adoption of cryptocurrencies by the mainstream population.