Welcome to insideAI News’s “Heard on the Street” round-up column! In this regular feature, we highlight thought-leadership commentaries from members of the big data ecosystem. Each edition covers the trends of the day with compelling perspectives that can provide important insights to give you a competitive advantage in the marketplace. We invite submissions with a focus on our favored technology topics areas: big data, data science, machine learning, AI and deep learning. Enjoy!
Impact of GPU/AI chip economics. Commentary by AB Periasamy, co-founder & CEO of MinIO
“GPU scarcity is going to continue to push AI workloads to the cloud. While most enterprises and AI companies would rather run in a private cloud, the reality is that they are at least six months from available supply to do so. As a result, they will take the economic hit associated with moving the data and compute to one of the major public clouds.
The silver lining here for 2024 is that the very GPU scarcity that is helping Nvidia achieve record revenues will drive software innovation and workload optimizations that will reduce future demand. This should have the effect of bringing down GPU pricing in the back half of 2024. Coupled with the arrival of competitive GPU offerings from AMD (a certainty) and Intel (very likely) may further depress prices and enable the next generation of workloads to run in the private cloud – away from prying eyes.
The key will be in the software ecosystem. The savvy analyst knows that Nvidia’s chips are only half the story – the other half is Cuda. The open source community has some work to do to create an equivalent software stack – but there is power in numbers and 2024 will see the playing field level out on the GPU front – both on the hardware and software side.
Once GPUs become more available and people understand those cloud economics, many will repatriate workloads to on-prem/private cloud where they can truly optimize and accelerate AI innovations.“
AI’s Pandora’s Box Is Already Open: Here’s How to Manage It. Commentary by K. Scott Griffith CEO and managing partner, SG Collaborative Solutions, LLC
“Artificial Intelligence is taking the world by storm, ushering in a new age of productivity and well-being. But AI also can cause great harm, both deliberate and accidental. For example, studies have shown AI imaging models demonstrate bias against historically underserved demographics in disease diagnosis. Or when a recent AI-generated legal brief referenced a fictional case in court. If left unconstrained, this harm will be irreversible. But if we wait for regulators to address the problem, we’ll be waiting for decades. We must take responsibility for harnessing AI’s positive potential, while doing more than we’ve done in the past to safeguard against harm.
What’s the solution? A strategic, well-organized government/industry/labor collaboration is needed. An example of such a collaboration is the U.S. airline industry’s Commercial Aviation Safety Team (CAST), which has led to a 95% reduction in the fatal accident rate. This success depended on the FAA’s role as the independent oversight body protecting the public; industry’s understanding of the technical challenges; and labor’s role as the eyes and ears of the evolving risk in frontline operations. At a time when aviation was at the forefront of advancing automation technologies, CAST was able to successfully harness knowledge, skills, and abilities to manage emerging risks. The industry moved from being reactive to proactive, and plane crashes in the US became a rarity.
We have the same opportunity to manage AI. Input and support must come from legislators, law enforcement, the research and educational communities, and consumer advocacy organizations. Private businesses must lead, instead of waiting until compliance becomes mandatory and burdensome. Regulators must move from compliance to risk-based oversight and labor must be incentivized to actively participate with economic and social benefits. Transparency and trust are key, while aligning legislation and the legal systems to better address the rapidly advancing challenge.”
The developing database market. Commentary by Percona founder Peter Zaitsev
“As vector data becomes more mainstream, dedicated, specialized vector databases are emerging in hopes of satisfying the growing demand. But it’s important to keep in mind that these systems offer highly specialized capabilities at the exclusion of many other, equally important ones. That’s why, at the same time, we’re beginning to see solutions that aim to integrate vector search and other vector capabilities within mainstream databases. Whether through integrations or extensions, I expect that, for the lion’s share of enterprise users, these types of solutions will offer more cohesive, lightweight, and familiar solutions to their AI development needs.”
ChatGPT’s latest Updates: Safety First, Always. Commentary by Mohammad, Founder, writerbuddy.ai
“OpenAI has taken a cautious approach while introducing capabilities to understand voice and image, which is highly commendable. They are clear about the model’s limitations and have plans to avoid potential misuse of information, such as avoiding identifying human faces. They proved that innovation can also coexist with ethical considerations, as they should.”
Generative AI: Failure to Prepare is Preparing to Fail. Commentary by Kevin Campbell, CEO, Syniti
“In a tech landscape abuzz with the potential of generative AI, understanding its power and pitfalls is a must. Organizations across every vertical are looking into how to leverage this technology to get ahead or simply ensure they can keep up with the latest and greatest advancements. Rushing implementation and forgoing the necessary learning process isn’t going to be worth the investment. A thoughtful, concerted approach that is rooted in data is essential for generative AI. To truly reap the advantages of AI investments, organizations must take a close look at their data strategy and ensure they are prepared for success. If using it to solve specific business problems is the goal, it’s key to consider how the training process for those models is done and examine the underlying foundation of data that you’re supplying in order to get the result you need.
Data is what trains generative AI models. To get the results most aligned to the business outcomes you’re hoping to gain, quality and governed data is the key component. The adage “garbage in, garbage out” applies here. If you have the right data sets that you want to use to train a model or gain insights, but the quality is poor, then you’re still going to end up with a bad result. Poor-quality data can lead to problems like inaccurate recommendations and irrelevant guidance. Generative AI holds immense promise for potentially transforming numerous industries. However, rushing into generative AI without a robust data strategy can lead to costly mistakes and delays. Data quality and context are the linchpins of success. Building strong data foundations and refining AI models iteratively are essential. A thoughtful approach, rooted in data, will ensure that organizations harness the full potential of generative AI, driving meaningful business outcomes.”
Successful DataOps Requires People, Process, and Technology in 2024 Now More than Ever. Commentary by Amit Patel, Senior Vice President at Consulting Solutions
“As people (data stewards) and processes (business rules, data standards) mature, technology can keep data quality on track.
Processes create the foundation of a robust DataOps framework, enabling organizations to confidently navigate an increasingly complex data landscape. Technologies such as AI/RPA can identify areas where data intake and maintenance are dropping the quality—be it due to manual data entry, EDI, or other API—and can automatically rectify them using available services like Dun & Bradstreet. Such tools allow the profiling and monitoring of trends to understand anomalies that inject poor data quality and help fix them.
Finally, ML tools get organizations to the next level of data maturity by detecting patterns and making predictions. This predictive power allows DataOps teams to get ahead of quality issues and ensure that data continually improves.”
A Fresh Approach: Leveraging End-User Computing Data to Enhance Business Performance. Commentary by Amitabh Sinha, CEO & Co-Founder of Workspot
“The productivity of any employee can be significantly impacted by the quality of their end user experience. There are three challenges: (i) Do you know if your users are happy? Individual applications, like Teams and Zoom, ask the question about quality of experience, but that data isn’t available to IT. Some companies have deployed Digital Experience tools to understand the overall compute experience; (ii) Can you monitor happiness continuously? User happiness can go from good to poor with one update to the operating system, driver, new security agent, or an application update; (iii) How do you perform root cause analysis? Once you see a trend for one group of users how do you determine the root cause and solve it?
Modern Virtual Desktop Infrastructure (VDI) technologies collect end-user survey results, performance metrics and application updates for every session and user. IT teams can quickly identify unhappy users, and then use performance metrics to gain insights into various conditions impacting end-user satisfaction, such as CPU or memory usage, network conditions, round-trip time, and more. This allows IT teams to promptly address issues at both individual and group levels.
The ability to connect performance data to the end-user experience enables businesses to make informed decisions, proactively resolve issues, and fine-tune their computing environments. With the right tools, businesses can gain deep visibility into how end users perceive their computing environments, helping users achieve optimal productivity while maintaining security and agility.”
GenAI now copilots pharma research. Commentary by Hanjo Kim, SVP of global strategy and head of medicinal chemistry at Standigm
“Lab automation, a long-standing industry practice, has evolved from merely automatic ‘executions’ to include automatic ‘controls’ and ‘adaptations’ in the era of AI-powered research. This necessitates impeccable data quality, further emphasizing lab automation’s role in maintaining data integrity.
As science progresses from molecular to organism levels, integrating AI models across various stages of drug discovery requires high data compatibility and substantial contextual data—both strong drivers for lab automation. Scientists are actively seeking innovative systems to intelligently complete missing values in multidimensional data matrices. In addition, the advancement of personalized medicine and better diagnostics has produced numerous compounds in smaller quantities. Ensuring data reproducibility from the early stages of research, even in medicinal chemistry labs, is integral to this process.
Many researchers have also relied on tools like ChatGPT for writing research papers. The scientific community is keen to explore its potential in their daily routines and its capacity to create new methods for executing previously unimaginable lab tasks.
The advent of AI technology is revolutionizing scientific methods in every aspect—it’s not a matter of if but when.”
Rethinking preventative healthcare with AI and population data. Commentary by Samantha Roushan, SVP, Clinical Transformation at Cohere Health
“The latest advancements in artificial intelligence (AI) are inviting a new wave of healthcare delivery transformation, moving towards a more proactive, patient-centric approach than ever before. Notably, the healthcare industry is projected to nearly double its budget allocation for AI and machine learning technologies, increasing from 5.7% in 2022 to 10.5% in 2024. This shift is particularly crucial given the prevalence of chronic illnesses, which affect 60% of U.S. adults and account for 90% of healthcare expenditures. Moreover, this shift is important due to the lackluster adoption of value-based care models designed to proactively address health, which is due in part to the absence of such supportive technologies.
Responsibly employed AI holds the potential to revolutionize healthcare, offering proactive, data-driven care from an initial diagnosis to the post-treatment stage, with extensive and impactful benefits. Responsible AI excels in data processing, resulting in more precise diagnoses, personalized treatment plans, and proactive condition management, all of which can improve patient outcomes. The technology enables early detection and diagnosis by identifying subtle patterns in datasets and facilitating timely intervention.
Predictive care is another advantage, as responsible AI anticipates disease progression and potential complications, enabling swift intervention and efficient allocation of resources. It also guides both healthcare providers and patients in making informed care decisions, focusing on personalized treatment approaches and connecting patients with suitable physicians and care programs. Furthermore, responsible AI enhances communication within the healthcare system, fostering seamless information sharing among providers, thereby improving coordination and ensuring patients receive care at the right time.
Lastly, responsible AI modernizes and streamlines prior authorization processes by creating evidence-based care paths for patients, mitigating delays and enhancing the quality of care. By harnessing health data, responsible AI is poised to transform healthcare, making it more proactive, preventative, efficient, and, ultimately, centered on individual well-being.”
AI’s ability to create personalization in an overworked industry. Commentary by Austin Jordan, Head of Data Science, AI and ML, at Apixio
“The healthcare industry generates a vast amount of data — 80% of which is unstructured and difficult for providers and health plans to leverage to improve patient outcomes. Generative AI could help solve this data problem with its ability to synthesize complex datasets and provide actionable insights.
When applied in healthcare, generative AI could streamline workflows and reduce administrative burdens by creating concise patient summaries for clinicians before and after visits. With generative AI’s ability to follow specific instructions combined with clinicians’ expertise, it also has the potential to improve the patient experience through tailored education. Instead of receiving a generic, boilerplate instruction manual for their treatment plans, patients could receive personalized, AI-generated instructions, educational videos, and summaries that are unique to their needs and health history.
We’re not there quite yet. Generative AI still makes mistakes and needs to be monitored to prevent bias and ensure data security — especially in healthcare, where the stakes are so high. We need to be careful that we don’t rush to use AI unsupervised in areas where mistakes can be catastrophic, like recommending the wrong dosage or frequency of a medication. But this new reality could be within reach sooner than many realize. AI has improved faster than many believed possible. If the algorithm’s output can be safeguarded via clinical expertise and introduced into healthcare in a way that supports clinicians, not replaces them, healthcare will become more personalized, effective, and accessible.”
Sign up for the free insideAI News newsletter.
Join us on Twitter: https://twitter.com/InsideBigData1
Join us on LinkedIn: https://www.linkedin.com/company/insidebigdata/
Join us on Facebook: https://www.facebook.com/insideAI NewsNOW
