%20.png?width=860&height=430&name=AI%20Webinar%202%20(August%2016)%20.png)
In the ever-evolving landscape of AI, one domain stands out at the forefront: Generative AI. Its swift advancements exert a profound influence on present and future business success, thus necessitating a discussion about Strategizing for AI Disruptions. In a recent insightful webinar hosted by NuBinary, industry leaders JP Rosevear and Ali Sharifi delved into critical aspects of AI adoption, harnessing LLMs, and vendors, and effectively using cutting-edge tools. The webinar profoundly illuminated how enterprises can strategically wield these advancements to gain a competitive edge and secure enduring relevance.
What Does AI Allow?
As briefly explained in our first webinar, Understanding Generative AI and the Disruption in the Next 3-5 Years, on the significant shifts in the computing paradigm, it is evident that over the years, there has been a transition from personal computers to cloud computing and mobile devices. Importantly, the new AI paradigm shifts is changing both how users interact with computers and how businesses operate; moving from command-based interaction to intent-based outcomes. Nevertheless, the whole realm of AI poses challenges that businesses need to navigate.
Challenges In AI Adoption
Gleaning from this recently published chart by KPMG where, Ali drew a comparison regarding AI adoption rates in Canada, the USA, and the EU. Though the adoption levels could be better, they are hindered by some challenges AI poses, thereby causing a lag. Some of the challenges stem from news-induced fear, companies’ and governments’ restriction, and even bans on AI tools such as Chat GPT.
There are three groups of these concerns:
Group 1: Security, Privacy, Copyright, and Maintenance Concerns.
Group 2: Safety Bias, Censorship, Emotion, and Transparency (these affect individuals directly)
Group 3: Cost and Resources, Lack of Understanding, Fear of Job Displacement, Integration challenges, and Cultural resistance (organizational concerns)
Security is more important nowadays as the AI pipeline has introduced a new attack surface. Therefore, it is essential to exercise care to ensure the expected outcome of the AI models is achieved. Currently, attackers are targeting different levels of the AI pipeline, specifically data-related ones. An example of this is the prompt injection attacks on ChatGPT. Business must implement different levels of data security and model security at different levels of the AI pipeline.
Focus on Security and Privacy
Governments and cities were identified as the central issue, and this was addressed from two perspectives: their slow adoption of and regulation of technology. Highlighting these worries, it is crucial for the swift tech adoption and regulation by governments due to AI’s rapid disruption.
Amid these concerns, here are some identified regulations emerging to address security; Canada’s AIDA, US AI Bill of Rights, EU AI Act, ISO/IEC FDIS 5338, ISO/IEC AWI 27090, and ISO/IEC WD 27091. For privacy, the regulations are GDPR (General Data Protection Regulation) and CCPA (California Consumer Privacy Act).
Among these regulations, GDPR was recommended as the most suitable for AI implementation. This is because of its successful integration with AI principles. This alignment is vital as AI intersects with human rights considerations, an area where The EU Act has made great strides by putting humans as the center of attention.
Safeguarding Copyright
This section delved into how businesses and individuals can safeguard copyright and legal rights. This has been extensively discussed in our recent blog post, “Plagiarism and Copyright Battles in Generative AI.”
Given these copyright cases, it was summarized that there is currently a lack of clear and definitive regulations outlining permissible actions and restrictions.
Ensuring AI Code Maintainability
While many are familiar with result-yielding AI models lacking clear explanations, the hurdle lies in their upkeep. Unlike traditional code, AI model errors are intricate, comprising numerous parameters instead of simple lines of code.
Limited internal visibility worsens the problem, as data scientists, without software development training, need developer collaboration. Also, the concept of “model drift” surfaces in AI models like GPT-4, where answers to consistent questions can fluctuate daily due to frequent model updates, raising trust concerns in AI outputs.
The recommendation was companies should adopt software development best practices, including collaborating with developers, utilizing behavior-visibility tools, and comprehensive documentation, to tackle these challenges and improve code maintainability.
How Do We Leverage AI Now?
The speakers further emphasized the ongoing transformative shift in AI interaction for businesses and individuals. They urged enterprises to recognize and adopt this trend, as nearly all companies will be affected. This can be achieved through the proactive exploration of AI’s potential (i.e. through experimentation and testing, even if a complete overhaul isn’t imminent).
Leveraging Through Human Augmentation
This is a unique concept separate from advanced or global AI, which focuses on enhancing productivity. Formerly time-consuming tasks, spanning days to hours or hours to minutes, are revolutionized. Key domains are content creation (where large language models eliminate the tedious aspects, such as grammar and style edits, enabling the generation of diverse content like video game narratives or music) and research (by providing insights into new topics, summarization of new topics, and generating and validating new ideas).
Businesses and individuals can leverage this concept in two ways:
Direct Use
Businesses can leverage these tools and toolings right now through the two types of large language models (LLMs); Base LLMS and Instruction Based LLMs. It is key to understand how to navigate and apply tools such as ChatGPT, Bing, and DALLE-2, which help generate preferred content, from text to images, by inputting various instructions.
Vendor Integrations
During tool development or acquisition, businesses need to factor in vendor integration. Key steps include conducting vendor assessment, looking for an AI roadmap, understanding their approach (narrow vs broad), and seeking clarity on security and privacy measures to safeguard your data, as previously highlighted.
Some tools that have great examples of these integrations are MS Office – Copilot, Photoshop (Generative Fill), Hubspot (CRM), Webflow (Websites), Otter.ai (Meeting Note Taking), and in Software Development (understanding how hired vendors integrate tools such as GitHub-Copilot in technology development)
Generative AI in the Near Future
Imminent trends to watch for encompass the following domains:
Constructive (Destructive) Actions
There is the Instacart example as a Read-only (passive interaction). However, the upcoming shift involves over 800 plugins taking active roles, such as making orders or bookings, potentially raising privacy and security concerns. This will boost productivity as AI efficiently executes tasks using pre-existing knowledge.
Data Ingestion & Augmentation
The limited word capacities like ChatGPT’s 32,000 words will see a shift soon. Emerging technologies will be offering higher limits (e.g., 100,000 words). By this, it will enable businesses to leverage detailed knowledge for improved business workflow, consistency, and performance.
Multi-Modal
This involves integrating diverse communication channels into a single interface. For instance, OpenAI’s March demo of converting a website sketch into functional code hints at its potential for generating diverse content from images and text inputs.
Agents
This point is both interesting and scary at the same time. Owing to the ability to autonomously take non-read-only actions based on intent, without human permission. Despite its futuristic nature, prototypes like the Auto-GPT Pizza ordering experiment are already surfacing.
Conclusion
In conclusion, these developments hold immense potential, while simultaneously presenting significant challenges from a security perspective. The rapid progress of large language models indicates that what seemed distant just a year ago is now within reach, possibly even within three years. With these changes on the horizon, there’s a world of possibilities to explore. Countless techniques, tips, and strategies await discovery, even by the creators themselves. It’s time to start considering the implications for your business. Identify how to harness these innovations, and how to navigate the associated risks. The future is knocking—will you answer the call?
How can NuBinary Help with Your Company’s AI Adoption?
NuBinary can help your company with AI strategy and through the entire software development lifecycle. Our CTOs are capable of figuring out the best technology strategy for your company. We’ve done it several times in the past, on a big range of different startups. We have all the necessary knowledge and tools to keep your development needs on track. This can be either through outsourcing or hiring internal teams.
Contact us at info@nubinary.com for more information or book a meeting to meet with our CTOs here.
About the Speakers:
JP Rosevear, co-founder and CTO at NuBinary has an extensive 25-year career in the industry, with a background spanning startups to large enterprises. JP has held leadership roles in esteemed companies like Nova (Polar), Novell, and Mozilla. Thus, giving him a wide range of experience in the tech field. His work as a Fractional CTO has led to successful collaborations with numerous startups. An example is PUSH, which successfully exited to WHOOP in 2021. He builds high-performing teams and nurtures scalable, and repeatable technology development processes. JP’s insights are invaluable for businesses aiming to innovate and grow.
Ali Sharifi, also a co-founder and CTO at NuBinary, is a distinguished tech executive boasting over 15 years of extensive experience in the realms of security and machine learning. He has a strong track record in ensuring compliance and shaping the future of technology. Ali’s work is globally recognized. His role at NuBinary involves providing cutting-edge solutions to companies’ technical and operational needs. His expertise has helped organizations achieve regulatory compliance and drive innovation. Notably, Ali’s academic contributions as a Professor and Principal Investigator have significantly influenced the development of various fields. This ranges from security and privacy to machine learning and sustainability measurement.
With their combined expertise, JP Rosevear and Ali Sharifi guided an engaging discussion on generative AI. Thereby painting a comprehensive picture of its evolution, current state, and future.
