Copyright © 2023 IDG Communications, Inc.

It’s highly unlikely that you’ve missed the buzz surrounding generative AI, and specifically large language models (LLMs) like ChatGPT. In exact months, these have been hot courses everywhere, from social media to the news to everyday conversations, and we’ve only just begun to learn what generative AI could be capable of.

Generally speaking, gen AI refers to a category of machine learning (ML) techniques that can create content like images, music and text that closely resembles human-created content. LLMs, on the other hand, are neural networks with billions of parameters that have been trained on vast amounts of text data, which enables them to understand, process, and generate human-like language.

Together, these technologies offer a diverse range of applications that hold the potential to reshape diverse industries and amplify the quality of interactions between humans and machines. By exploring these applications, business owners and enterprise decision-makers can gain valuable inspiration, drive accelerated growth and achieve tangibly improved results through rapid prototyping. The added advantage of gen AI is that most of these applications require minimal expertise and do not require further model training.

Quick disclaimer: People often tend to associate gen AI exclusively with ChatGPT, but there are numerous models from other providers available, like Google’s T5, Meta’s Llama, TII’s Falcon, and Anthropic’s Claude. While most of the discussed applications in this article have made use of OpenAI’s ChatGPT, you can readily adapt and switch the underlying LLM to align with your specific compute budget, latency (how fast you need your model to generate completions — smaller models allow quicker loading and reduce inference latency), and downstream task.

1. Connect LLMs to external data

LLMs demonstrate impressive capabilities at many tasks right out of the box, such as translation and summarizing , without requiring initial customization. The reason they are so good at these generic tasks is that the underlying foundation model has been trained on large yet generic datasets. However, this competence might not seamlessly extend to domain-specific tasks including, for example, providing answers about your company’s annual report. This is where Retrieval Augmented Generation (RAG) comes into the picture.

RAG is a framework for building LLM-powered systems that make use of external data sources. RAG gives an LLM access to data it would not have seen during pre-training, but that is necessary to correctly provide relevant and accurate responses. RAG enables language models like ChatGPT to provide better answers to domain-specific questions by combining their natural language processing (NLP) abilities with external knowledge, mitigating instances of generating inaccurate information or “hallucinations.” It does so by:

• Retrieving relevant information from external knowledge sources, such as large-scale document collections, databases or the internet. The relevance is based on the semantic similarity (measured using, say, cosine similarity) to the user’s question.
• Augmenting the retrieved information to the original question in the prompt (to provide a helpful context for answering the question) and passing it to the LLM so it can produce a more informed, contextually relevant, and accurate response.

This approach makes LLMs more versatile and useful across various domains and applications, including question-answering, content creation and interactive conversation with access to real-time data. Podurama, a podcast app, has leveraged similar techniques to build its AI-powered recommender chatbots. These bots adeptly suggest relevant shows based on user queries, drawing insights from podcast transcripts to refine their recommendations.

This approach is also valuable in crisis management. PagerDuty, a SaaS incident response platform, uses LLMs to generate summaries of incidents using basic data such as title, severity or other factors, and augmenting it with internal Slack data , where responders discuss details and share troubleshooting updates to refine the quality of the summaries.

While RAG may appear intricate, the LangChain library offers developers the necessary tools to implement RAG and build sophisticated question-answering systems. (In many cases, you only need a single line of code to get started). LangChain is a powerful library that can augment and enhance the performance of the LLM at runtime by providing access to external data sources or connecting to existing APIs of other applications.

When combined with open-source LLMs (such as Llama 2 or BLOOM), RAG emerges as an exceptionally potent architecture for handling confidential documents. What’s particularly interesting is that LangChain boasts over 120 integrations (at the time of writing), enabling seamless functionality with structured data (SQL), unstructured content (PDFs), code snippets and even YouTube videos.

2. Connect LLMs to external applications

Much like utilizing external data sources, LLMs can establish connections with external applications tailored to specific tasks. This is particularly valuable when a model occasionally produces inaccuracies due to outdated information. For example, when questioning the present Prime Minister of the UK, ChatGPT might continue to refer to Boris Johnson, even though he left office in late 2022. This limitation arises because the model’s knowledge is fixed at its pretraining period and doesn’t encompass post-training events like Rishi Sunak’s appointment.

To address such challenges, LLMs can be enhanced by integrating them with the external world through agents. These agents serve to mitigate the absence of internet access inherent in LLMs, allowing them to engage with tools like a weather API (for real-time weather data) or SerpAPI (for web searches). A notable example is Expedia’s chatbot, which guides users in discovering and reserving hotels, responding to queries about accommodations, and delivering personalized travel suggestions.

Another captivating application involves the automatic labeling of tweets in real-time with specific attributes such as sentiment, aggression and language. From a marketing and advertising perspective, an agent connecting to e-commerce tools can help the LLM recommend products or packages based on user interests and content. 

3. Chaining LLMs

LLMs are commonly used in isolation for most applications. However, recently LLM chaining has gained traction for complex applications. It involves linking multiple LLMs in sequence to perform more complex tasks. Each LLM specializes in a specific aspect, and they collaborate to generate comprehensive and refined outputs.

This approach has been applied in language translation, where LLMs are used successively to convert text from one language to another. Companies like Microsoft have proposed LLM chaining for translation services in the case of low-resource languages, enabling more accurate and context-aware translations of rare words.

This approach can offer several valuable use cases in other domains as well. For consumer-facing companies, LLM chaining can create a dynamic customer support experience that can enhance customer interactions, service quality, and operational efficiency.

For instance, the first LLM can triage customer inquiries and categorize them, passing them on to specialized LLMs for more accurate responses. In manufacturing, LLM chaining can be employed to optimize the end-to-end supply chain processes by chaining specialized LLMs for demand forecasting, inventory management, provider selection and risk assessment.

Prior to the emergence of LLMs, entity extraction relied on labor-intensive ML approaches involving data collection, labeling and complex model training. This process was cumbersome and resource-demanding. However, with LLMs, the paradigm has shifted. Now, entity extraction is simplified to a mere prompt, where users can effortlessly query the model to extract entities from text. More interestingly, when extracting entities from unstructured text like PDFs, you can even define a schema and attributes of interest within the prompt.

Potential examples include financial institutions which can utilize LLMs to extract crucial financial entities like company names, ticker symbols and financial figures from news articles, enabling timely and accurate market analysis. Similarly, it can be used by advertising/marketing agencies for managing their digital assets by employing LLM-driven entity extraction to categorize ad scripts, actors, locations and dates, facilitating efficient content indexing and asset reuse.

5. Enhancing transparency of LLMs with ReAct prompts

While receiving direct responses from LLMs is undoubtedly valuable, the opaqueness of the black box approach often raises hesitations among users. Additionally, when confronted with an inaccurate response for a complex query, pinpointing the exact step of failure becomes challenging. A systematic breakdown of the process could greatly assist in the debugging process. This is precisely where the Reason and Act (ReAct) framework comes into play, offering a solution to these challenges.

ReAct emphasizes on step by step reasoning to make the LLM generate solutions like a human would. The goal is to make the model think through tasks like humans do and explain its reasoning using language. One can easily operationalize this approach as generating ReAct prompts is a straightforward task involving human annotators expressing their thoughts in natural language alongside the corresponding actions they’ve executed. With only a handful of such instances, the model learns to generalize well for new tasks.

Taking inspiration from this framework, many ed-tech companies are piloting tools to offer learners personalized assistance with coursework and assignment and instructors AI-powered lesson plans. To this end, Khan Academy developed Khanmigo, a chatbot designed to guide students through math problems and coding exercises. Instead of merely delivering answers upon request, Khanmigo encourages thoughtful problem-solving by walking students through the reasoning process. This approach not only helps prevent plagiarism but also empowers students to grasp concepts independently.

Conclusion

While the debate may be ongoing about the potential for AI to replace humans in their roles or the eventual achievement of technological singularity (as predicted by the godfather of AI, Geoffrey Hinton), one thing remains certain: LLMs will undoubtedly play a pivotal role in expediting various tasks across a range of domains. They have the power to enhance efficiency, foster creativity and refine decision-making processes, all while simplifying complex tasks.

For professionals in various tech roles, such as data scientists, software developers and product owners, LLMs can offer valuable tools to streamline workflows, gather insights and unlock new possibilities.

Varshita Sher is a data scientist, a dedicated blogger and podcast curator, and leads the NLP and generative AI team at Haleon.

Microsegmentation is table stakes for CISOs looking to gain the speed, scale and time-to-market advantages that multicloud tech stacks provide digital-first business initiatives.

Gartner predicts that through 2023, at least 99% of cloud security failures will be the user’s fault. Getting microsegmentation right in multicloud configurations can make or break any zero-trust initiative. Ninety percent of enterprises migrating to the cloud are adopting zero trust, but just 22% are confident their organization will capitalize on its many benefits and transform their business. Zscaler’s The State of Zero Trust Transformation 2023 Report says secure cloud transformation is impossible with legacy network security infrastructure such as firewalls and VPNs. 

Defining microsegmentation

Microsegmentation divides network environments into smaller segments and enforces granular security policies to minimize lateral blast radius in case of a breach. Network microsegmentation aims to segregate and isolate defined segments in an enterprise network, reducing the number of attack surfaces to limit lateral movement. 

It’s considered one of the main components of zero trust and is defined by NIST’s zero-trust framework. CISOs tell VentureBeat that microsegmentation is a challenge in large-scale, complex multicloud and hybrid cloud infrastructure configurations and they see the potential for AI and machine learning (ML) to Boost their deployment and use significantly.

Gartner defines microsegmentation as “the ability to insert a security policy into the access layer between any two workloads in the same extended data center. Microsegmentation technologies enable the definition of fine-grained network zones down to individual assets and applications.”  

Microsegmentation is core to zero trust 

CISOs tell VentureBeat that the more hybrid and multicloud the environment, the more urgent — and complex — microsegmentation becomes. Many CISOs schedule microsegmentation in the latter stages of their zero-trust initiatives after they’ve achieved a few quick zero trust wins. 

“You won’t really be able to credibly tell people that you did a zero trust journey if you don’t do the micro-segmentation,” David Holmes, Forrester senior analyst said during the webinar “The time for microsegmentation is now,” hosted by PJ Kirner, Illumio cofounder and advisor. 

Holmes continued: “I recently was talking to somebody [and]…they said, ‘The global 2000 will always have a physical network forever.’ And I was like, “You know what? They’re probably right.’ At some point, you’re going to need to microsegment that. Otherwise, you’re not zero trust.”

CIOs and CISOs who have successfully deployed microsegmentation advise their peers to develop their network security architectures with zero trust first, concentrating on securing identities often under siege, along with applications and data, instead of the network perimeter. Gartner predicts that by 2026, 60% of enterprises working toward zero trust architecture will use more than one deployment form of microsegmentation, up from less than 5% in 2023. 

Every leading microsegmentation provider has active R&D, DevOps and potential acquisition strategies underway to strengthen their AI and ML expertise further. Leading providers include Akamai, Airgap Networks, AlgoSec, Amazon Web Services, Cisco, ColorTokens, Elisity, Fortinet, Google, Illumio, Microsoft Azure, Onclave Networks, Palo Alto Networks, Tempered Networks, TrueFort, Tufin, VMware, Zero Networks and Zscaler.

Microsegmentation vendors offer a wide spectrum of products spanning network-based, hypervisor-based, and host-agent-based categories of solutions.

How AI and ML simplify and strengthen microsegmentation

Bringing greater accuracy, speed and scale to microsegmentation is an ideal use case for AI, ML and the evolving area of new generative AI apps based on private Large Language Models (LLMs). Microsegmention is often scheduled in the latter stages of a zero trust framework’s roadmap because the large-scale implementation can often take longer than expected. 

AI and ML can help increase the odds of success earlier in a zero-trust initiative by automating the most manual aspects of implementation. Using ML algorithms to learn how an implementation can be optimized further strengthens results by enforcing the least privileged access for every resource and securing every identity.   

Forrester found that the majority of microsegmentation projects fail because on-premise private networks are among the most challenging domains to secure. Most organizations’ private networks are also flat and defy granular policy definitions to the level that microsegmentation needs to secure their infrastructure fully. The flatter the private network, the more challenging it becomes to control the blast radius of malware, ransomware and open-source attacks including Log4j, privileged access credential abuse and all other forms of cyberattack.

Startups jumping into the space

Startups see an opportunity in the many challenges that microsegmentation presents. Airgap Networks, AppGate SDP, Avocado Systems and Byos are startups with differentiated approaches to solving enterprises’ microsegmentation challenges. AirGap Networks is one of the top twenty zero trust startups to watch in 2023. Their approach to agentless microsegmentation shrinks the attack surface of every connected endpoint on a network. Segmenting every endpoint across an enterprise while integrating the solution into a running network without device changes, downtime or hardware upgrades is possible.

Airgap Networks also introduced its Zero Trust Firewall (ZTFW) with ThreatGPT, which uses graph databases and GPT-3 models to help SecOps teams gain new threat insights. The GPT-3 models analyze natural language queries and identify security threats, while graph databases provide contextual intelligence on endpoint traffic relationships. 

Prime areas for AI and ML

AI and ML can deliver great accuracy, speed and scale in microsegmentation in the following areas:

Automating policy management

One of the most difficult aspects of microsegmentation is manually defining and managing access policies between workloads. AI and ML algorithms can automatically model application dependencies, communication flows and security policies. By applying AI and ML to these challenges, IT and SecOps teams can spend less time on policy management. Another ideal use case for AI in microsegmentation is its ability to simulate proposed policy changes and identify potential disruptions before enforcing them.

More insightful, real-time analytics

Another challenge in implementing microsegmentation is capitalizing on the numerous sources of real-time telemetry and transforming them into a unified approach to reporting that provides deep visibility into network environments. Approaches to real-time analytics based on AI and ML provide a comprehensive view of communication and process flows between workloads. Advanced behavioral analytics provided by ML-based algorithms have proven effective in detecting anomalies and threats across east-west traffic flows. These analytics Boost security while simplifying management.

More autonomous asset discovery and segmentation

AI can autonomously identify assets, establish communication links and identify irregularities and distribute segmentation policies without manual intervention. This self-sufficient capability diminishes the time and exertion needed to execute microsegmentation and maintains its currency as assets alter. It additionally mitigates the potential for human error in policy development.

Scalable anomaly detection

AI algorithms can analyze extensive amounts of network traffic data, allowing for the identification of abnormal patterns. This empowers scalable security measures while maintaining optimal speed. By harnessing AI for anomaly detection, microsegmentation can expand across extensive hybrid environments without introducing substantial overhead or latency. This ensures the preservation of security effectiveness amidst the expansion of the environment.

Streamlining integration with cloud and hybrid environments

AI can Boost microsegmentation’s integration across on-premises, public cloud and hybrid environments by identifying roadblocks to achieving optimized scaling and policy enforcement. AI-enabled integration provides a consistent security posture across heterogeneous environments, eliminating vulnerabilities attackers could exploit. It reduces operational complexity as well.

Automating incident response

AI allows for automated responses to security incidents, reducing response times. Microsegmentation solutions can use trained ML models to detect anomalies and malicious behavior patterns in network traffic and workflow in real-time. These models can be trained on large datasets of normal traffic patterns and known attack signatures to detect emerging threats. When a model detects a potential incident, predefined playbooks can initiate automated response actions such as quarantining affected workloads, limiting lateral movement and alerting security teams. 

Enhanced collaboration and workflow automation

AI streamlines team collaboration and automates workflows, decreasing the time required for planning, analysis and implementation. By enhancing collaboration and automation, AI has optimized the entire microsegmentation lifecycle, allowing for a quicker time-to-value and ongoing agility, thereby enhancing the productivity of security teams.

Essential to zero trust architecture 

Microsegmentation is essential to zero trust architecture, but scaling it is difficult. AI and ML show potential for streamlining and strengthening microsegmentation in several key areas, including automating policy management, providing real-time insights, enabling autonomous discovery and segmentation and more. 

When microsegmentation projects are delayed, AI and ML can help identify where the roadblocks are and how an organization can more quickly reach the results they’re after. AI and ML’s accuracy, speed and scale help organizations overcome implementation challenges and Boost microsegmentation. Enterprises can reduce blast radius, stop lateral movement and grow securely across complex multicloud environments.

How can businesses facilitate learning and development on a shoestring budget? originally appeared on Quora: the place to gain and share knowledge, empowering people to learn from others and better understand the world.

Answer by Susan Anderson, Chief Services Officer at Mineral, on Quora:

The learning and development gap is widening between enterprise organizations and small- to mid-sized businesses. According to Mineral’s 2022 State of HR Survey, 84% of large businesses report their employees engage in regular training, compared to just 58% of small businesses. How does this gap play out in the real world? Just this year, Microsoft flexed its L&D investment muscle by creating a fully-produced TV series – Trust Code – exclusively for its employees and strictly for the purpose of promoting compliance training. Meanwhile, only 47% of small organizations increased their investment in HR compliance last year. Of course, we could never expect a small business to produce a TV series to Boost its training efforts, but there are substantive steps SMBs can take to develop their teams without breaking the bank.

Before we dive into the best practices for budget-friendly L&D, why should SMBs care so much about training? For one thing, it’s the most reliable way to retain your employees. Recent research suggests that the Great Resignation not only isn’t over, but it’s more of a long-term trend than we initially thought. That argument should be convincing enough to free up some L&D budget, especially when you consider nearly all US employees (94%) agree they are willing to stay longer at a job that invests in their professional development.

Another good reason for SMBs to care about training comes from the exact Supreme Court decision on Affirmative Action. The ruling that colleges and universities must stop considering race in admissions could lead to a less diverse pool of job candidates graduating with higher education degrees. Nearly 70 employers – including Google, Johnson & Johnson, Starbucks and Uber – stated in a brief to the Supreme Court that the absence of Affirmative Action could cause them to lose a “pipeline of highly qualified future workers and business leaders.” That doesn’t have to mean a less diverse workforce, however. Organizations that want to continue promoting diversity may need to increase their emphasis on skills-based hiring over degree-based hiring. While degree-based hiring may now exacerbate systemic inequities post-Affirmative Action, skills-based hiring widens the talent pool and can inject a ready-to-learn mentality into the workplace. As degrees become less important, on-the-job training will become more important.

Now that you’re (hopefully) convinced of the importance of training, how can budget-constrained SMBs get the most training bang for their buck?

Invest in 3rd party training

In a exact webinar on employee training, Steven T. Hunt – Chief Expert at SAP and author of the best-selling book “Talent Tectonics” – noted that technology is removing one of the biggest development hurdles that used to face SMBs. It used to be that a small business’ inability to create an in-house Learning Management System (LMS) meant they couldn’t train their employees as well as enterprise organizations. Now, online learning libraries have democratized the access to high quality training, such that SMBs can now afford much of the same training content as their larger counterparts. Rather than investing in a massive suite of training materials with categories they don’t need, SMBs can invest in a right-sized LMS solution through a third-party provider.

The best bet for an SMB to find a solution that fits their budget is to only pay for the training packages they need. For most organizations the hierarchy of learning needs – where training for job basics and compliance represent the base of the learning pyramid – will provide a good reference point. The more appetite for training an organization has (and the more budget), the higher on the pyramid they will go – to more advanced needs at the top of the pyramid, such as culture development and leadership skills. Most organizations will pay to train for the basics of the job, which could also be described as job-specific skills training. Considering the increased importance in skills-based hiring, this is a crucial place for any SMB to start. Compliance training may not look as sexy at an SMB as it will at Microsoft, but your organization doesn’t want to foot the bill for a compliance violation. Not only will compliance training greatly help out your HR department – 68% of HR departments reported that maintaining compliance was “a very time-consuming effort” in last year’s Mineral State of HR Survey – but it is also required in several states. The most budget-conscious organizations will start there. As they grow, the right LMS partner can help them scale their training offerings to meet their evolving needs.

A great way to supplement these up-front investments is to deliver training in a group setting where participants can discuss key learnings and apply scenarios to their specific business. Free training resources like Coursera or Alison don’t come with certifications but can also help businesses supplement training.

Generative AI

If leveraged properly and with thoughtful precautions, Generative AI can be a powerful tool for creating and developing a training program. There are two perfect situations that would position an organization to do this effectively, both of which involve collaboration between a human expert and an AI tool. Situation #1 involves an organization that employs in-house experts on key training topics, but that doesn’t have a dedicated L&D professional on staff. A company may, for example, have an expert in Leadership Skills who has no background in L&D and thus is not proficient at creating a training program. The expert could collaborate with Generative AI to build a Leadership curriculum, ideally leveraging a library of preferred content on the subject. The AI could read and use this Leadership content as the basis for creating learning materials and curriculum flows. Generative AI can even create quizzes and answer keys based on instructed parameters.

Situation #2 is the exact opposite – an organization has a skilled L&D professional but doesn’t employ an expert in a key training subject. A skilled L&D practitioner can effectively leverage AI to design and deliver meaningful training for a subject in which they are less familiar. Perhaps your company’s L&D director needs to develop training for new safety protocols. While they aren’t familiar with the subject, they could use AI as a means of gaining functional or technical knowledge on the subject before they develop a training program.

Regardless of your situation, if you choose to leverage AI for training, it is imperative that you are thorough in their research. Carefully craft the prompts you feed to the AI tool and cross-reference any answers you receive before developing training to guard against potential misinformation. It is also best practice to follow the ADDIE (Analyze, Design, Develop, Implement, Evaluate) framework – a recognized best practice methodology for developing training.

Tap Into Thought Leaders

Training resources abound online. By leveraging insights from thought leaders via webinars and podcasts, organizations can access best practices and resources on courses like emerging technologies, customer engagement, culture building and beyond. The key is first to find thought leaders that are trustworthy and that you find engaging, and to make sure you are always mining for the most up-to-date information available. Yesterday’s best practices could change tomorrow, which makes it important to find a range of experts whose content you can source. Speaking of best practices, it’s definitely a good idea to supplement this strategy with other training materials, unless you want to spend all your time searching for your team’s next webinar.

Leverage in-house team members

An uncomfortable reality in every business is turnover – it is inevitable. While eliminating it is impossible, there are development strategies that can help position your organization to fill roles quickly and effectively when someone leaves. Promoting job rotation and cross-training is a particularly effective method, allowing each member of an organization to learn different skills by rotating roles. This practice helps develop a universal understanding of the business and can also position employees for promotion to new roles in the event of turnover.

In a similar vein, teaching others can be the best way to master and share skills. SMBs can empower their team to choose different courses to research and organize teach-back sessions to build expertise within the organization. These groups should regularly rotate topics, allowing everyone to experience the material as both a student and a teacher. This is also a great way for organizations to Boost engagement in their training. Not only is it a safe learning environment, but it also enables inquisitive learners to use their own research to fill gaps in knowledge from previous teachers, creating the potential for thoughtful dialogue on each subject.

This question originally appeared on Quora - the place to gain and share knowledge, empowering people to learn from others and better understand the world.

Succeeds Jim Milton, who will be retiring

Dubai, United Arab Emirates: Anthology, a leading provider of education software solutions that support the entire learner lifecycle, announced today that Chief Executive Officer and Chairman Jim Milton will retire after more than nine years of service to the company. He will be succeeded by Bruce Dahlgren, who has been appointed incoming CEO.

During his time as CEO, Milton oversaw several transformational acquisitions. In 2020, Anthology emerged as a new company through the combination of higher education technology leaders: Campus Management, Campus Labs, and iModules. In 2021, the company merged with Blackboard, creating a comprehensive EdTech ecosystem able to deliver its higher education clients deeper data, analytics, and insights to drive student and institutional success.

“It has been an honor to lead this company in support of the thousands of clients and millions of users that we serve through our world class solutions,” said Jim Milton. “I am confident Anthology will continue to grow its influence and impact in higher education. Anthology is courageously innovative, and I am excited to see what’s next for the company under Bruce’s leadership.”

Bruce Dahlgren, a seasoned technology company executive with over three decades of expertise in the B2B software space, was most recently the CEO of MetricStream, the global leader in enterprise cloud platform and applications for integrated Governance, Risk, and Compliance. At MetricStream, he led a new go-to-market growth strategy and delivered best-in-class SaaS metrics, overseeing double-digit annual contract value (ACV) growth. As CEO of Anthology, Dahlgren will focus on accelerating the company’s expansion in cutting-edge technologies, delivering high quality client support and a superior client experience, supporting the company’s leading partnership relationships, and further integrating technologies across the company’s user base in more than 80 countries.

“Jim and the talented team at Anthology have created the only holistic software ecosystem for the higher education community - one that has the power to transform the learner journey and help institutions deliver on their missions,” said Bruce Dahlgren. “I am honored to be joining this innovative and dedicated team to continue Anthology’s work of delivering purpose-built solutions to institutions and improving learner outcomes around the world.”

“We thank Jim for his thought leadership, focus on innovation and unwavering commitment to Anthology clients,” said Ramzi Musallam, CEO and Managing Partner of Veritas Capital, which owns Anthology. “We are confident that Bruce will seamlessly carry forward these values and leverage his proven expertise to lead Anthology on its continued mission of transforming lives through the power of education.”

Dahlgren’s appointment comes at a time of significant growth for Anthology. Last month, the organization announced the incorporation of generative AI capabilities into its ecosystem of EdTech solutions through their long-standing collaboration with Microsoft. The company also announced new product features that are facilitated by AI and aimed at improving student success and retention rates.

About Anthology

Anthology offers the largest EdTech ecosystem on a global scale for education, supporting more than 150 million users in 80 countries. With a mission to provide dynamic, data-informed experiences to the global education community through Anthology Intelligent Experiences™, we help learners, leaders and educators achieve their goals by offering over 60 SaaS products and services designed to advance learning. Discover more about how we are fulfilling our mission for education, business and government institutions at www.anthology.com.

Copyright © 2023 IDG Communications, Inc.

SAN FRANCISCO, Aug. 17, 2023 — DynamoFL, Inc. announced that it has closed a $15.1 million Series A funding round to meet demand for its privacy- and compliance-focused generative AI solutions. Coming off a $4.2M seed round, the company has raised $19.3M to date. DynamoFL’s flagship technology, which allows customers to safely train Large Language Models (LLM) on sensitive internal data, is already in use by Fortune 500 companies in finance, electronics, insurance and automotive sectors.

The round, co-led by Canapi Ventures and Nexus Venture Partners, also had participation from Formus Capital, Soma Capital and angel investors Vojtech Jina, Apple’s privacy-preserving machine learning (ML) lead, Tolga Erbay, Head of Governance, Risk and Compliance at Dropbox and Charu Jangid, product leader at Snowflake, to name a few.

The need for AI solutions that preserve compliance and security has never been greater. LLMs present unprecedented privacy and compliance risks for enterprises. It has been widely shown that LLMs can easily memorize sensitive data from its training dataset. Malicious actors can exploit this vulnerability to extract sensitive users’ personally identifiable information and sensitive contract values with carefully designed prompts, posing a major data security risk for the enterprise. The pace of innovation and adoption in the AI sector is punctuated by the rapidly changing global regulatory landscape, many of which require that enterprises detail these data risks, but enterprises today are not equipped to detect and address the risk of data leakage. In the EU, the GDPR and the impending EU AI act, along with similar initiatives in China and India, as well as AI regulation acts in the US, require that enterprises detail these data risks. However, today they are not equipped to detect and address the risk of data leakage.

More clearly needs to be done. As government agencies like the FTC explore concerns around LLM providers’ data security, DynamoFL’s machine learning privacy research team recently showed how personal information – including sensitive details about C-Suite executives, Fortune 500 corporations, and private contract values – could be easily extracted from a fine-tuned GPT-3 model. DynamoFL’s privacy evaluation suite provides out of the box testing for data extraction vulnerabilities and automated documentation to ensure enterprises’ LLMs are secure and compliant.

“We deploy our suite of privacy-preserving training and testing offerings to directly address and document compliance requirements to help enterprises stay on top of regulatory developments, and deploy LLMs in a safe and compliant manner,” said DynamoFL co-founder Christian Lau.

“Privacy and compliance are critical to ensuring the safe deployment of AI across the enterprise. These are foundational pillars of the DynamoFL platform,” said Greg Thome, Principal at Canapi. “By working with DynamoFL, companies can deliver best-in-class AI experiences while mitigating the well-documented data leakage risks. We’re excited to support DynamoFL as they scale the product and expand their team of privacy-focused machine learning engineers.”

The company’s solutions help organizations privately fine-tune LLMs on internal data while identifying and documenting potential privacy risks. Organizations can choose to implement DynamoFL’s end-to-end suite or to implement their Privacy Evaluation Suite, Differential Privacy and/or Federated Learning modules individually.

DynamoFL was founded by two MIT PhDs who have spent the last six years researching the cutting-edge, privacy-focused AI and ML technology forming the basis of the company’s core offerings. The team balances expertise in the latest research in privacy-preserving ML, with researchers and engineers from MIT, Harvard and Cal-Berkeley, and experience in deploying enterprise-grade AI applications for Microsoft, Apple, Meta and Palantir, among other top tech companies.

“This investment validates our philosophy that AI platforms need to be built with a focus on privacy and security from day one in order to scale in enterprise use cases,” said DynamoFL CEO and co-founder Vaikkunth Mugunthan. “It also reflects the growing interest and demand for in-house Generative AI solutions across industries.”

“While AI holds tremendous potential to transform every industry, the need of the hour is to ensure that AI is safe and trustworthy. DynamoFL is set to do just that and enable enterprises to adopt AI while preserving privacy and remaining regulation-compliant,” said Jishnu Bhattacharjee, Managing Director, Nexus Venture Partners.”We are thrilled to have partnered with Vaik, Christian and team in their journey of building an impactful company.”

About DynamoFL, Inc.

DynamoFL is the world’s leading enterprise solution for privacy-preserving Generative AI. At DynamoFL we believe that prioritizing privacy, compliance and data security from day 1 while building Generative AI applications is the only way to responsibly scale AI and use it to augment human potential beyond what was thought possible. Our proprietary technology encapsulates state of the art optimization techniques for training and deploying Generative AI models along with a robust privacy training and evaluation suite incorporating paradigms like Federated Learning and Differential privacy to bring high performing end-to-end plug-and-play Generative AI to global enterprises.

Source: DynamoFL

AI is everywhere. It’s in our apps, it’s on our smartphones, it’s developing new strains of neural know-how within the confines of the cloud datacenter and it’s rolling out across the ‘edge’ compute estate of smart devices in the Internet of Things (IoT). Because Artificial Intelligence (AI) is so ubiquitous, it is also various, multifarious and occasionally precarious (when we fail to eradicate AI bias and explainability) in its nature. Today, the biggest driving force in AI comes from the fact that it is now capable of working in ways that deliver not just predictive intelligence, but also generative intelligence.

We’re talking about generative-AI of course.

While we see every enterprise technology vendor on the planet now work to deliver a degree of this new smartness in its platform, applications and tools, it is compelling to look into the mechanics and infrastructure behind its delivery. Why is this so? Because this is not just an oil change, this is an engine refit in many senses i.e. while an AI accelerator can be a simple turbo-charge for some applications, many deployments of this technology will require data workload management at the infrastructure level in order for the technology to work to its full potential - or in some cases to work at all.

AI infrastructure first, smart apps second

Having spent the majority of its 25-year history working to provide IT infrastructure choice across storage services, networking and application management & virtualization, as well as being a key player in the cloud infrastructure management space, VMware is now continuing its systems level development by working with Nvidia to deliver core services that underpin new AI deployments. VMware and Nvidia have expanded their existing partnership to help enterprises that run on VMware’s cloud infrastructure to be ready for the era of generative AI.

While we normally understand the term ‘foundation’ to refer to some type of institution, enterprise IT firms sometimes like to use it to denote a base-level framework competency (Microsoft did it with Windows Communication Foundation back in 2006). Using that same style of naming protocol, VMware Private AI Foundation with Nvidia is designed to enable enterprises to customize foundational models (a technology we have explained here) and run generative AI applications. This could be smart apps that might include chatbots, intelligent assistants, search and summarization services - the latter being a way of using AI to categorize and filter masses of information that might exist in emails, for example. In this case, we see a platform that will exist as an integrated product featuring generative AI software and accelerated computing from Nvidia, built on VMware Cloud Foundation and optimized for AI.

“Generative AI and multi-cloud are the perfect match,” said Raghu Raghuram, CEO, VMware. “Customer data is everywhere — in [company] datacenters, at the [IoT] edge and in their clouds. Together with Nvidia, we’ll empower enterprises to run their generative AI workloads adjacent to their data - with confidence - while addressing their corporate data privacy, security and control concerns.”

Aligning AI adjacency

That point noting ‘adjacency’ to data’ is important. Speaking to press and analysts on a video call this week, VMware’s Paul Turner, vice president of product management vSphere and cloud platform echoed Raghuram’s sentiment by explaining how and why this adjacency is a reality.

“One of the things we believe companies will do now is to bring more of their generative AI workloads to their data, versus moving their data into public cloud services,” said Turner. “These same company’s may run some form of generative AI services with cloud service providers [in more public arenas], but we believe quite a lot of companies and a lot of major enterprises will want to run these technologies in relatively small [more restricted] environments. That way they protect themselves and they protect their data, which is their key asset.”

Jensen Huang, founder and CEO of Nvidia backs up the central messages being delivered here and says that, in the ‘race’ to integrate generative AI into businesses, his firm’s expanded collaboration with VMware will offer organizations the full-stack software and computing they need to unlock the potential of generative AI using custom applications built with their own data. All well and good so far then, but we wanted to know more about how these new strains of generative AI will be adequately supported at the infrastructure level.

A difference in inference

We know that the true power of generative AI happens when it can apply the scope of Large Language Model (LLM) data assets and produce human-like levels of inference i.e. this is intelligence that creates contextualized 'things' that have been inferred from an understanding of the other information around them. Talking about this area and how his firm's Graphical Processing Units (GPUs) work to accelerate the speed at which this intelligence is delivered, Justin Boitano, vice president of enterprise computing at Nvidia explained that his firm's latest Bluefield 3 GPUs deliver 1.4 times extra performance for generative AI on the inference side, more in some cases too.

"As we all know now, corporate data is the new asset, so how you manage your data, how you optimize your data, how you take models like LLaMA and bring that model to colocate within your data in your datacenter, all matters a lot," said Boitano. "We're seeing great innovation in this space [with technologies like pre-training, fine-tuning and in-context learning] to optimize generative AI tuning so that it is relevant to each and every business and is able to create new business value offerings. We're seeing this insight in VMware where we're looking at auto encoders so that we can take our API's and our SDKs, then feed them through an automation mechanism driven by Nvidia - and we're able to actually generate pretty good code samples. It needs further work. Of course, you need to then work on optimizing the model, but the capability and the capacity is there," he noted, in the same call to press and analysts.

As we move ahead here then, what do VMware and Nvidia think the key enabling technologies and identifying functions will be? For certain, analyst house McKinsey estimates that generative AI could add up to US$4.4 trillion annually to the global economy. Looking again at the technologies on offer here, VMware Private AI Foundation with Nvidia is designed to enable organizations to customize essentially open data based Large Language Models; produce more secure and private models for internal usage; and offer generative AI as-a-Service to users.

All of which, on a paper and in practice, will lead to an ability to securely run ‘inference workloads’ (the computing guts that delivers new human-like AI) at major scale. The platform is expected to include integrated AI tools to deliver customers what VMware calls ‘proven models’ trained on their private data in a cost-efficient manner. Being finalized now and rolled out through 2024, the technology here is built on VMware Cloud Foundation and Nvidia AI Enterprise software.

According to the Nvidia team, “The platform will feature Nvidia NeMo, an end-to-end, cloud-native framework included in Nvidia AI Enterprise — the operating system of the Nvidia AI platform — that allows enterprises to build, customize and deploy generative AI models virtually anywhere. NeMo combines customization frameworks, guardrail toolkits, data curation tools and pretrained models to offer enterprises an easy, cost-effective and fast way to adopt generative AI.”

Why infrastructure is rising

This story circulates around the central ways that generative AI is being enabled at the infrastructure level. Because VMware is also delivering functions as automations and assistants for network engineers and developers (the company likes to define its audience into platform teams, networking teams and end user teams), via natural language, the use of these technologies can also arguably broaden.

It’s a wider democratization of technology trend that VMware chief technology officer (CTO) Kit Colbert has explained very clearly. “The line between applications and infrastructure has changed. Things that used to be considered infrastructure (Kubernetes for cloud container orchestration is a good example) have now become infrastructure. Why? Because of the inherent standardization that has happened to make technologies at this level usable and popular in the first place,” said Colbert. “So now, what we must realize is, the infrastructure line itself is always rising.”

We can dovetail these thoughts with other new products from VMware. The company has now introduced a suite of technologies across VMware Tanzu, its modular cloud-native application development, delivery and operations optimization and visibility platform. Because Tanzu is modular and is underpinned by a common data platform and control with support for open interfaces, it enables broad ecosystem integrations. This is multi-cloud management technology that works with VMware’s own Aria product, a multi-cloud management portfolio for managing the cost, performance, configuration and delivery of infrastructure and applications.

“Tanzu and Aria are now evolving into the next generation of Tanzu Application Platform and the new Tanzu Intelligence Services. With an application-centric focus and integration through common data and controls, VMware Tanzu is providing a streamlined platform engineering and cloud operations experience and better software agility,” said Purnima Padmanabhan, senior vice president and general manager, modern apps and management business group at VMware.

Padmanabhan explains that VMware is announcing Tanzu Application Platform to now combines new innovations for platform engineering and operations with the existing capabilities of Tanzu for Kubernetes operations to help companies deliverwhat she calls a ‘world-class’ internal platform.

“Managing applications across clouds is a web of data and technology complexity. Distributed silos of tools and data make it difficult to gain visibility into the dependencies between applications, infrastructure and services. Centralizing management of these disparate systems and enabling shared data helps eliminate silos. This empowers teams to respond more quickly to issues and to continuously tune applications and environments using deep and actionable insights,” notes Padmanabhan and team, in a technical statement.

What is VMware now?

All of which developments analyzed and offered here hopefully clarify some of how the backroom engines running the new breed of generative AI (and indeed, old fashioned predictive AI) will work.

Does that mean VMware is becoming a company that will start to now offer generative AI applications and services in a tangible sense?

No, says CEO Raghuram… and VMware probably wouldn’t ever want to anyway i.e. it wants to do what it has always done which is to provide a competent and all encompassing infrastructure offering that enables firms to always have choice across server, networks, applications, cloud and now Large Language Models in the world of generative AI. It’s a logical enough progression and there is a large product portfolio here ‘underpinning’ this technology proposition - pun not intended, but useful nonetheless - it all starts with infrastructure.