Here’s what’s on our plate today:

• ☁️ Cloud muscle behind AI, datacenters, decides who wins the race.

• 🏀 AI scouts NBA talent, researchers track pro moves for bot training.

• 🍔 Uber Eats goes visual and snaps a menu pic, gets dish details.

• 🖥️ Female-led Chip Upstart, SixSense, scores fresh funding.

Let’s dive in. No floaties needed…

The Laboratory

Why cloud systems are the real game-changers driving AI adoption

Scientific breakthroughs have always driven world-changing technologies. Each leap forward prompts businesses and societies to reorganize and capitalize on new possibilities. For the past few centuries, this phenomenon has played out repeatedly as humans evolved from horse-collar-era workflows to the 21st century, when computers reshaped how we view work.

During the last technological shift, tech companies changed the way businesses and users viewed the PC (Personal Computer), paving the way for the information revolution. This time around, Artificial Intelligence (AI) is changing the nature of knowledge-based work, from coding to writing to creating speech and video. However, behind the scenes, a quieter transformation is enabling AI’s integration into everyday workflows: the shift to cloud computing.

Where does computing power dwell?

Since the introduction of the computer, whenever there was a need for more computing power, enterprises and individual users had to upgrade their hardware. However, as workflows grew, so did the need for more powerful computer systems, which could not be housed within a single device. This led to the creation of the cloud system, large data centers in remote locations housing information that could be accessed without the need for a physical connection, thanks to advancements in telecommunications. With AI, the cloud has become an essential part of workflows because it provides the computational power, scalability, storage, and accessibility that modern systems require.

The role of cloud computing in AI workflows

Training modern AI models, especially large language models (LLMs) or image recognition systems, requires massive GPU or TPU clusters. Cloud providers like AWS, Google Cloud, Microsoft Azure, and Oracle Cloud offer on-demand access to such infrastructure. For example, OpenAI’s models like the GPT-4 are trained and deployed using Microsoft’s Azure AI supercomputing infrastructure. Azure provides the necessary GPUs, high-speed networking, and storage to train models with hundreds of billions of parameters.

For users, regardless of their workflow, implementing AI systems via the cloud is an attractive option as it removes the traditional barriers to entry, like cost, infrastructure, and expertise. For Small and Medium Enterprises (SMEs), cloud computing reduces the burden of investing in expensive on-premise hardware and allows them to access powerful AI tools and infrastructure on a pay-as-you-go basis through cloud platforms like AWS, Google Cloud, and Microsoft Azure.

An example of this was seen when an Indian startup, Infinity Learn, partnered with Google to harness the latter’s AI stack to deliver hyper-personalized learning experiences. Google, in a blog post, shared that more than 60% of the world’s generative AI startups have been building on Google Cloud, including Anthropic, Cohere, Magic, Mistral, and AI21 Labs.

The benefits of AI cloud computing are showcased by the growing number of organizations opting for it. According to a report from TechTarget Enterprise Strategy Group, three-fourths of organizations worldwide are running their GenAI workloads on public cloud providers.

However, even as smaller enterprises are looking to AI companies to handle AI workflows in the cloud, they have to contend with challenges that are unprecedented.

The challenges of cloud AI systems

One of the major challenges for enterprises is understanding their needs and opting for the best option for them. This starts with the question of whether they should opt for the public cloud computing model offered under the multi-tenancy model adopted by major cloud providers or invest in creating private cloud infrastructure.

Enterprises handling Personally Identifiable Information (PII) have to comply with many data privacy regulations. AI projects increase the risks because massive amounts of real data dictate the behavior of ML training models. Additionally, compliance with local legal requirements around data residency and geolocation can be expensive and complicated.

Another area of concern is ensuring data security. While cloud providers offer robust security, breaches continue to disrupt workflows. In June 2024, Mandiant identified a threat campaign targeting Snowflake customer database instances with the intent of data theft and extortion. The company processes proprietary and sensitive data of many Forbes Global 2000 companies on AWS, Azure, and Google Cloud.

Finding the right talent is another challenge. Cloud computing is a relatively new development, and as such, finding personnel with expertise and experience can be challenging. Additionally, finding data scientists, AI and ML engineers, and prompt engineers are roles that are added when implementing AI systems. All these are difficult roles to fill.

Uncertainty around return on investment

A key concern around investments in AI workflows is their return on investment. Right now, investments are focused on building out the raw computing power, electricity infrastructure, and communication setup. This had led to a massive impact on the stock prices of enterprises working on them. However, for small enterprises, while the cost of implementing AI workflows remains high, enterprises remain unsure of what the return on their investments in AI might look like.

So far, the return on investment, when it comes to AI, has been playing out in the stock market and has largely centered around the performance of the so-called Magnificent Seven: Amazon, Apple, Alphabet, Meta Platforms, Microsoft, Nvidia, and Tesla.

And while around 75% of SMEs are at least experimenting with AI, with growing businesses leading in adoption (83%),  according to a blog post from Salesforce. Returns are yet to be measured.

What’s the future outlook?

With the growing adoption of AI systems to supplement workflows and, in some cases, replace them, we are witnessing the early stages of a larger shift. It may take years before the early adopters get to show a return on their investment. As for the users and enterprises still on the fence, a staggered and cautious approach may be the right choice. Such an approach can include adopting Modular AI systems, partnering with AI‑focused startups and accelerators, and investing in qualified workers, or upskilling existing staff via online courses, workshops, or industry programs, which will help ease the transition and prepare for the larger shifts in workflows.

Roko Pro Tip

Bite-Sized Brains

• AI Researchers Train Like NBA Stars — Labs are borrowing pro-basketball regimens—film study, advanced metrics, and ‘practice squads’—to sharpen model performance and teamwork.

• Uber Eats lets AI read the Menu — Snap a pic, get dish descriptions, allergens, and crowd-sourced reviews in seconds; rollout starts with US restaurants this fall.

• SixSense Scores Fresh Funding — Woman-led chip startup raises $45 M to build ultra-efficient AI accelerators, aiming to dethrone Nvidia in edge computing.

Quick Poll

Meme Of The Day

Rate This Edition