Ali Ghaemi

DeepSeek AI

Jan 292025

AI/Artificial intelligence has been a hot topic ever since OpenAI released its ChatGPT over two years ago. A Chinese competitor, called DeepSeek, has made waves with its own offering this week. The app went viral becoming the #1 downloaded app on Google Play Store and Apple’s App Store.

Observers marveled at a product that reportedly matched or surpassed its Western competitors in several benchmarks in a short time – despite using less advanced Nvidia chips. The Biden administration banned the export of Nvidia’s top-of-the-line chips to China to impede that country’s AI ambitions.

While some publications cautioned users against using DeepSeek’s app due to data privacy concerns, it’s important to note that privacy issues are not unique to Chinese companies. The U.S. government has access to data from American tech companies through various laws and programs.

Let’s address several misconceptions:

While DeepSeek did admit to using secondary and less powerful chips, it did not deny having previously acquired and deployed Nvidia’s top chip, the H100. In fact, many reports pointed to DeepSeek having amassed an inventory of the same prior to the ban going into effect.
The experts advocating for non-usage were also seemingly off the mark. We already know through the NSA revelations and public knowledge laws, like the Patriot Act, that privately generated data entrusted to US tech companies is shared with the US Government and, from there, other allied intelligence agencies. Many point out that those are not equivalencies. Fair enough. Nonetheless, it does bear pointing out as all the stories I read omitted the fact.
Moreover, does anyone remember Cuil? The search engine by two former Google search employees launched sometime in 2008 and immediately received wide press coverage including for its main thrust that it had the most indexed pages of any search engine, Google included. Cuil was everywhere seemingly for a few weeks, yet shut down and went out of business in 2010. That is not to say that DeepSeek will meet the same fate as Cuil. It just serves as a reminder that the race is a marathon and not a sprint and proof should be in a pudding older than a week.

In the world of AI data is everything. Consistently acquiring and synthesizing it is the determinant of an AI’s success – as far as technology goes, that is – as the myriad of models out there all seemingly work fine. Yet, given the ongoing data ingestion requirements a winner is never assured. In fact, as of today another Chinese company Alibaba’s Qwen-2.5, beats Anthropic’s Claude 3.5 Sonnet, Google Gemini-2 Flash and OpenAI’s GPT-4o in most benchmarks. It also surpassed DeepSeek’s results. It is important to note that DeepSeek R1 used other available datasets to achieve its operation rather than going out there and ingesting the universe of all data. This also implies that DeepSeek is subject to the same hallucinations as its progenitors.

Finally, and most importantly, for the viability of AI and the universe of GenAI users, it is imperative to point out the bottom-line: DeepSeek is good news for all AI including the more established competitors. How? Firstly, it brings AI further into the public consciousness. The articles and publicity bring GenAI to a wider audience and one that is global. Secondly, let’s remember that DeepSeek was put to work serving its founder’s (Liang Wenfeng) main occupation, namely the High-Flyer hedge fund. By all accounts, the man and his hedge fund have been wildly successful thanks to the power of artificial intelligence, which is now in your, and my, hands. Can anyone think of a better validation and advocacy for AI technology and its myriad of possibilities and use cases?

Things That Need To Go Away: Judging AI by news cycles, rather than focusing on its engagement and market share capture across various use cases and verticals.

AI Transformation: Implementation Across Industries

Articles Add Comments »

Jan 192025

Generative Artificial Intelligence is no longer a futuristic concept. It is a business imperative. As technologies evolve at (even more) unprecedented speeds, organizations across industries need practical strategies to harness AI’s potential. The below provides business leaders and technology managers with a blueprint for understanding, evaluating and implementing AI technologies that drive value.

A post by Sam Altman, CEO of OpenAI, reported last week that the company is losing money on its top tier ChatGPT Pro, which was launched at $200 per month in December 2024. The tier offers users unlimited access to OpenAI’s latest model, OpenAI o1, as well as access to its Sora AI video generator and Advanced Voice Mode. In other words, in case it was not obvious, the more subscribers, the bigger the loss. There is also the possibility that Altman managed to slide in an ad for OpenAI’s top tier of course.

There is also a free tier and the limited access Plus model for $20 per month.

It was refreshing that the CEO accepted responsibility for the pricing and the loss, which he did. OpenAI had projected losses of about $5 billion (US) and revenue of $3.7 billion for 2024 as of last summer. Amazingly, in one recent report from Jfrog ML 87% of AI users said they were leveraging OpenAI’s model. It highlights how market share could come with financial challenges. Some may recall that Amazon lost money for years before becoming what it is today.

Assuming that $200 per month is a ceiling for what most people will pay for access, and knowing that the company is still losing money, it is reasonable to deduce that the company needs to make up for the shortfall and encourage uptake and usage elsewhere.

With that in mind, here is some help for Altman and co. to present the need for their solution and gain more customers and revenue. The use cases could apply to other providers like Cohere, DeepSeek-R1, Anthropic, AWS Bedrock or Google Vertex as well. Below is a summary of the utility and use cases for AI both in general and by vertical.

Recommendations for AI implementation

Start Small, Scale Smartly: Begin with a pilot project to demonstrate value before full-scale deployment.
Data Strategy: Emphasize the importance of a robust data management strategy, including data quality and governance.
Continuous Learning: Implement a feedback loop for model improvement and adaptation to changing needs.
Ethical Considerations: Incorporate ethical AI principles into the deployment process.

Steps and prerequisites to deploy AI:

Assess needs and corresponding applications.
Budget for and procure the technologies and resources needed for data ingestion, integration, testing and on-going optimization.
Ingest in-house, public or both types of data including structured, unstructured or both types of data.
Designate team members responsible for the model, policies and operation including the HITL (Human-In-The-Loop) model requirement.
Map and connect data with designated business outputs and desired outcomes.
Train employees and stakeholders.
Continuously monitor and review risks, security and evolving regulatory compliance (see above designated team requirement).

Mainstream and commercially available AI that could be deployed today could take different forms. The actual ones utilized are dependent on the use case. These include:

Analysis and didactic AI
Conversational and Personal AI
Developer AI
Image/Visual AI (or multi-modal AI potentially)
Predictive AI
Process AI
Search AI and
Task AI

Zendesk reports that “in a few years,” 80% of customer interactions will be powered by AI – a significant leap from the 20% we see today. There really is no time to waste then. In the same context, the fact that the tech has gone mainstream so recently means new use cases will emerge all the time.

Why AI?

Increased efficiency and productivity
Innovate more and faster
Improve customer and employee experience
Impact revenue and margins positively
Reduce costs
Maintain and improve competitive position
Manage risk, governance and compliance

General use cases:

The standard use cases for Generative AI are horizontal ones. These include:

Automating things,
Speeding up tasks like knowledge augmentation and management,
Software development,
Customer service and
Recommendations and personalization

More specifically, and beyond the above bullet points below is a summary of likely use cases by industry.

Industry	Broad Generative AI Utilities
Healthcare	– Medical image analysis and enhancement – Clinical data synthesis and analysis – Personalized treatment planning – Drug discovery acceleration – AI-assisted diagnostics and treatment planning
Financial Services	– Risk assessment and fraud detection – Personalized financial advice generation – Automated report and document creation – Market trend analysis and prediction – AI-driven compliance and regulatory reporting
Manufacturing	– Product design optimization – Predictive maintenance – Supply chain optimization – Quality control enhancement – AI-optimized energy management in production
Retail & Ecommerce	– Personalized product recommendations – Customer behavior prediction – Inventory management optimization – Dynamic pricing strategies – AI-powered visual search and product recognition
Technology & Software	– Code generation and optimization – Automated testing and debugging – User interface design – Natural language processing for user interactions
Media & Entertainment	– Content creation and editing – Personalized content recommendations – Audience engagement prediction – Virtual character and environment generation
Education	– Personalized learning content creation – Automated assessment and grading – Educational content summarization – Adaptive learning path generation
Real Estate	– Property valuation modeling – Virtual property tour generation – Market trend analysis – Client matching and recommendation
Energy & Utilities	– Energy demand forecasting – Grid optimization – Fault prediction and diagnosis – Renewable energy output prediction
Agriculture	– Crop yield prediction – Pest and disease detection – Precision farming optimization – Weather impact analysis

As the AI landscape evolves, companies must remain agile, focusing on value creation while managing costs effectively.

Things That Need To Go Away: Worrying About The Dangers Of AI As Reason For Inaction

HOW NOT TO SELL BY MIKE WICKS

Book Reviews Add Comments »

Jan 012025

Don’t Do This!

In my experience, salespeople are often cynical about sales advice, classes, tutorials and books by persons who have either never primarily worked in sales or haven’t been professional salespeople. That appears to be the case with the author of this book, Mike Wicks. His biography and parts of the book’s narration suggest he has some sales experience and has held a variety of roles that are related to the field, but he has never worked full-time in a sales job or held the official title of salesperson. On page 39, There is a story about being in sales and getting kicked out by the owner, resulting in the narrator, and his sales samples, landing in the gutter. Not sure whether the story is true, but if it is, it sounds both nasty and funny at the same time (incidentally, the narrative is about matching the personality of the buyer). Regardless, the aforementioned prejudice may, or may not, be valid. After all, prospects, users and folks who understand human behaviour surely have a lot to teach salespersons.

Nonetheless, the book did attract me from the get-go. It may be the indirect approach the title portends, which is original if nothing else, or perhaps that the book is a quick read. It could also be because my own posts always end with an antithesis. The book’s focus may come across as focusing on what to avoid doing, but this is always contrasted with what the right behaviour looks like. Each (short) chapter contains at least one story, which makes reading the book more enjoyable.

The book is divided into three sections: It’s All About You, It’s All About What You Know and It’s All About The Sale.

One reason the book is an easy read is that the 200 pages are not densely packed and the chapters are short. Another is that each chapter repeats itself, not only with the aforementioned antithesis, but also with an instant summary of what was just expanded on. In general, the advice is elementary and the book is better suited to less experienced sales professionals. Having said that, the chapters do become progressively more intricate and developed. To be clear, the advice is always relevant and accurate, but it would be hard to imagine a seasoned salesperson being unaware of it. Still, it is a good compendium and comprehensive reminder.

How Not To Sell reminds the reader that there are four types of personality. These are the Analytical, Drive, Analytical and the Amiable.

Analyticals make good accountants or scientists. They may be reserved and proceed methodically. Make sure your facts are correct when speaking with them and show you want to support their personal credibility. Sell to them by being thorough and precise. Do not rush them and bring evidence. Analyticals need technical specifications and details. Use pros and cons and reverse engineer the process.

Drivers make good CEOs or surgeons. Their concern is the bottom-line. They do not appreciate people who hesitate. They are all about their output and are competitive. Show them how you can help them achieve their desired results. Sell to them by focusing on the task and emphasizing the results. Be concise and answer “what” questions. Drivers prefer not to be bogged down by excessive details. Drivers need to know your product or service makes them money, saves time and does both quickly. Demonstrate how they can make money or save resources and deploy assumptive closes.

Expressives make good salespeople or facilitators. They are stylish and like to be the centre of attention. They are often popular and chameleon-like. Expressives tend to bore easily. Help build their standings with others and you’ll get on. Sell to them by focusing on the relationship and promote a conversation. They are not detail-oriented either and like approachable people. Expressives need the feel-good factor. Be their friend and assume you have the order.

Amiables make better teachers or nurses. They take things slowly and care for relationships. Sell to them by being relaxed and approachable. Amiables need logical and systematic approaches, but also predictable. Agree often and listen. Amiables need guarantees and knowing that they are protected. Build the relationship and recommendation close with ongoing support and a solid warranty in place.

The book has several good presentation tips that are worth noting. Here’s a hint: Keep it short, be clear how you can help them after demonstrating you understand the challenges they face, revisit the challenges at both the start and end and, for heaven’s sake, do not read your bullet points verbatim.

Lastly, do not forget to sell yourself, sell your company and sell your product!

It’s Coming, It’s Coming, It’s Here (A Tale Of Two AIs)

Articles 1 Comment »

Dec 072024

Absolutes are typically mythical, being analogous to sasquatch and E.T. Trends and developments are rarely linear.

There are usually pros and cons in everything. Benefits and hazards are built into most things. Even an accelerating race car hesitates momentarily before it gathers more speed. That is the nature of things.

Artificial Intelligence has done a lot of good and will continue to do a lot of good. It is not for nothing, however, that many of the smartest voices of the scientific community warn us about the dangers of AI. Many of these forebodings are unprecedented and cataclysmic. Somewhere in-between, and more imminent, is the tectonic shift to our day-to-day lives that is happening right in front of our eyes. Online education platform Chegg has lost half a million paid subscribers, its market cap has crashed from $15 billion (US) to $300 million and it has laid off 25% of its employees. What will this, for example, do to authors?

True Artificial Intelligence, not the imposter AI that everyone and anybody is touting these days, following the training phase learns independently, benefits from advanced machine learning algorithms and deep neural networks, such as deep learning, and improves without direct human input. Certain AI models, like transformer-based architectures (the ‘T’ in GPT) rely on data to improve their ability to reply and make (better) predictions overtime. This is what makes this technology different from anything that has come before it. It is self-reinforcing and can be unleashed to run independently. Change after change in history, including a variety of technologies, has moved the landscape, made certain skills and jobs redundant while creating others. Think of a bridge replacing a ferry. Computers replacing typewriters. Cloud-based software (SAAS) replacing dedicated workstations, among others. This one is different.

With that said, it would be both societal and personal mistakes to stay away and try to close the stable door after the AI-driven, robotic horse has bolted. Sticking one’s head in the sand is not the solution. More likely, not making an effort and attempting to learn passively is not adequate. It is counterproductive. The fact of the matter is, and not many would admit it, that AI is not well understood even by the scientists and programmers at some of the best-known tech organizations of today. In such an environment, the most sane way to proceed, is to counterintuitively lift the lid, bring AI to the masses, give everyone every opportunity to use the tech, be educated on it and make it publicly sourced. Legislation is desirable, and necessary, because guardrails and moats do not build themselves, but even more powerful is a public taught to understand what AI is, what it is not, what it does, what it does not do and transparently knows where and how this technology is deployed. Shining a light, embracing and understanding is the best antidote to ignorance and the best way to insulate oneself from becoming redundant thanks to AI – or any change for that matter.

Things That Need To Go Away: Being Scared Of AI And Trying To Dock

What Is Governance In AI?

Articles Add Comments »

Nov 282024

In an earlier post, I discussed the importance of governance in Artificial Intelligence (AI) and how, arguably, aside from the initial hurdle of getting started, governance is one of the most significant barriers to adoption, particularly in large enterprises. Concerns such as liability, intellectual property and the risk of introducing incorrect or biased information into AI models are often cited as the biggest impediments to AI integration at scale.

My previous advice encouraged experimentation, emphasizing the importance of gaining momentum, learning from efforts and celebrating small wins. However, as promised, this follow-up aims to define what governance in AI really means. The first paragraph above provides some context, but let’s dive deeper.

AI-governance

Governance in AI refers to the set of practices, principles and processes that an organization establishes to develop, deploy and manage its Artificial Intelligence systems. In practice, this encompasses all systems that provide data to the AI, all outputs and outcomes generated by the AI and all stakeholders – individuals, teams, departments – whose jobs, roles and successes are influenced by AI. Since AI is fundamentally built on data, this broad reach underscores the technology’s far-reaching impact.

AI is still relatively new in the wider society and not fully understood. It is imperative that the governance framework adopted by an organization is designed with a clear end-goal in mind, and is implemented transparently, with widespread knowledge across the organization. This approach helps the AI initiatives align with organizational acceptance.

This does not imply that organizations should become paralyzed by over-analysis, as failing to implement AI would likely mean falling behind in today’s competitive landscape. The key to success lies in balancing careful governance with agile action. Trust is a vital component of AI adoption, and proper governance fosters trust by ensuring transparency and accountability.

Additionally, AI systems must be regularly monitored and evaluated to ensure they continue to function as intended, without introducing unforeseen risks or biases. This ongoing governance is essential for maintaining the public’s trust in AI technologies, as well as ensuring compliance with evolving regulations.

AI governance is multifaceted, but definitely possible and practical. Keeping a human in the loop is a check against an unintended consequence. Diverse stakeholders need to focus on long-term goals and organizations must engage to harness the full potential of AI while minimizing risks and fostering trust.

Things That Need To Go Away: The ‘AI Can Wait’ Attitude

Data Governance In AI

Articles Add Comments »

Nov 132024

I speak with buyers on a mission to procure the right product and services for their project. They include decision-makers figuring out the best applications for their companies.

They are all interested in AI and range from experimenting with LLMs (Large Language Models) and ML (Machine Learning) in silos to those who are eager to unleash AI for all their employees to take advantage of a range of possibilities. More and more, every one of the above conversations conveys a simple concept: AI is becoming pivotal to everything we will do.

Did you know that 91% of info/tech companies have mentioned ‘AI’ in their earnings calls at least once so far this year (2024)? Yet, a common concern arises consistently in conversations with enterprise management:

They are concerned about AI governance.
The compliance team is worried about data integrity and inputs.
The security team is daunted by the task of implementing AI that impacts their products and
Senior executives are uneasy regarding the potential implications if anything goes wrong during customer and end-users interaction with the technology.

These are legitimate concerns.

We will address and define what governance in AI is in a subsequent post. For now, it is important to remember that there are no absolutes and, the answer is, that we all must have the same expectations of AI as we have with any other piece of technology. Put another way, it makes sense to think of AI in the same way we expect the Internet or SAAS applications to function. Having said that, there are actions that technology custodians can and must take.

For AI governance to be done right we need to meticulously follow responsible data practices. These include:

Documenting the origin of the data and educating the user base is a start.
Similar transparency in where the data will be deployed and which use cases it covers is required.
Abiding by applicable laws is a must and non-negotiable. These include the EU Artificial Intelligence Act, which went into force on the 1st of August 2024. Additional regulations are en route, from federal and state jurisdictions, such as Canada’s AI And Data Act. The legal framework includes the above bullet points, yet it is important to remember that the framework and implementation will always evolve, so keep an eye on its implementation and evolution.
Keep a Human In The Loop. In other words, designating a person to interact with the LLM ensures human oversight; allowing for timely intervention when needed. The underlying models are getting better and algorithms learn and improve, but HTL allows for human intervention in any case.

In tandem, we know that the LLMs by IBM, Meta and others allow for scrutiny and the peace of mind of the user community to some extent because they are open source in code and licensing. Other models strive to accomplish the same credibility by offering access to the foundation models. This does not infer perfection. It does imply scrutiny and a level of credibility, however.

Being concerned and diligent is warranted. Not moving forward due to fear of the technology, however, is a recipe for falling behind. To stay competitive, remain informed about AI developments, begin experimenting and consider a low-risk use case for an initial quick win.

Things That Need To Go Away: AI At Any Cost And No AI At Any Cost

A Call For Professionalism And Communication

Articles Add Comments »

Nov 012024

We have all read them – or worse, been there. These are accounts of people who had a job interview set up and the interviewer failed to show up to the initial telephone or video interview, dropped off the radar mid-process or when a recruiter went AWOL and didn’t provide any feedback at the end of the process.
I was reminded of this conundrum upon coming across an article about a new website called Ghostedd. Visitors are invited to “share a ghost story.” The merits and trustworthiness of the service aside, the question for me was why such a service is needed in the first place? This Guardian article, which was printed the same week as the aforementioned one, fuels the conversation by reminding readers of fake jobs (as much as 40%), which sadly remains an ongoing issue.

Call me naive, but why can’t we have a process where we respect one another’s time, humanity and need for feedback and information? Recently, I came across a LinkedIn post where a recruiter was empathizing with someone who, having been ghosted by a company during her interview process, was expressing frustration about the system. Ironically, the same recruiter had tried to recruit me some ten years ago and then failed to follow up, update or give feedback after executing a disappearing act. The rub? He had called me and insisted what a great fit it was and how the hiring vice-president was eager to meet. The follow-up and meeting he’d promised a week later never materialised. To be fair, the recruiter may have evolved over the past decade, but if not, then his comment highlights a need for greater honesty in addition to propriety.

The flip side, which is not a mystery to most, is that recruiters and hiring managers often feel hesitant. Short on time and also facing potential blowback or litigation if they were to speak their minds and give actual feedback, they take the easy and safe way out. Yet, surely, there is a fine line between ghosting someone, placing yourself in a compromising position or ignoring someone you’ve put in a pipeline process. Moreover, if candidates reciprocate the courtesy and consideration expected in the process and given to them then interviewers should feel more assured in giving feedback and not rudely abandoning those they have engaged. For now the Internet is doing what it was supposed to do by crowdsourcing information, but could we all agree that when we are more courteous and considerate of our fellow humans and make their journey better, everyone benefits and is in a better position? This should not be a revolutionary concept. After all, if nothing else, we may be the ones looking for proper etiquette or propriety next time.

Things That Need To Go Away: Decenit a.k.a. Lack Of Mutual Respect, Lack of Feedback, Lack Of Explanation, Lack Of Follow-up And Fake Job Postings

Elevate Festival

Articles Add Comments »

Oct 072024

The agenda for the Elevate Festival in Toronto caught my attention. The event took place at Meridian Hall and the St. Lawrence Centre for Arts, which are nearly adjacent venues, last week. The combination of sessions, speakers and, what they call, masterclasses seemed intriguing both personally and professionally. My primary focus is AI at the workplace and those sessions were my main reason for attending naturally.

First though, one question lingers for anyone who knows. There were neither floats nor confetti. No one was dancing, throwing candy or engaging in revelry, but seriously, the question was, why is it called a “festival”?

Elevate festival

Oddly, there was another ‘festival’ called Future Festival occurring in Toronto concurrently. Surely, it can’t be a coincidence.

Returning to the conference, while AI was a prominent theme, the three-day event offered more than just discussions on the hot topic of Artificial Intelligence. There were start-ups and funders, a sales course, sessions for creative professionals and implicit infomercials for the speakers and their companies. Met a couple of nice people in line looking to sell their services and found out outsourced marketing and sales departments are more in-demand than ever. Some say AI is going to make both obsolete… or is outsourcing departments the first step towards elimination?

Here are a few things learned from the event:

A speaker emphasized that AI is software and should be treated exactly as such. That is, it can be used for good or evil and we should have the same expectations for it as we have for (other) software, the Internet or other pieces of technology. Think about it, that is the gist of it, isn’t it?
Emphasis was placed on the importance of experimentation. AI is still in its infancy and enterprises need to move forward even when use cases are not fully defined and also because there will be many use cases that companies do not even know they need yet. It is important to begin experimenting and exploring possibilities.
A few humanitarian use cases cited and demonstrated were: wildfire detection, Foundation For Healthcare, which is like AI from a medical school, and Alphafold, which by classifying and identifying the structure of protein is advancing medicine in the fight against disease.
A speaker mentioned that whatever you do, don’t do the ostrich.
According to KPMG research, companies aiming to excel in SEO and demand generation for AI should aim to publish three pieces of content (such as a white paper, blog post or solution PDFs) on their websites each week.
While transparency and accountability are crucial in the usage and deployment of AI, most LLMs (Large Language Models) already incorporate ethical practices. This was done even before the EU’s Artificial Intelligence Act went into force on August 1st, mandating GDPR, privacy and PII practices.
The CDO (Chief Data Officer) from Telus and IBM’s VP for AI shared that they can’t be any more explicit about safety protocols built into AI than by saying they have opened AI to all their employees, including developers, internally themselves without exception.
IBM, Meta (which has experienced 350 million downloads of its Llama model) and Mozilla all reminded attendees that their LLMs are open source and subject to scrutiny by the world-wide community.
KPMG’s Canadian AI lead: “AI will be in every function in every industry” and “our analysis shows a 72% ROI in 3 three years” for the enterprise, numbers he characterized as “unchartered” return numbers.
Finally, someone boldly described the situation with AI a rising tide lifting all boats, highlighting the collective potential of these advancements.

The event was educational even if there was a lot of ‘what’ without the ‘how’ at the sessions on my agenda. Companies like Telus, Google, Meta, IBM and Mozilla made the effort to appear.

Things That Need To Go Away: ‘Educational’ Sessions At Paid Conferences That Are Implicit Advertisements

Artificial Intelligence (AI) And Sales & Related Jobs

Articles Add Comments »

Jun 092024

AI, in general, is a hot topic everywhere. This site has discussed the nature of AI before and posted about the hype versus reality of it. Every passing day has made the promise and reality of AI more real, more rewarding and more ominous. The aforementioned AI article listed several vendors who had begun working on inserting artificial intelligence and its brain, machine learning, into sales and its various departments.

A recent quotation by Eric Yuan, chief executive of Zoom, the video conferencing company, was worth attention. He is suggesting that an AI version of us can attend meetings for us in five or six years. The implication is that a digital version of us could act for us and, moreover, would be as diligent, acceptable and effective as the person it represents. Yuan was careful to couch this as an augmentation technology as opposed to a replacement one. The scenario again was the utopian dream of having much more leisure and downtime. Yet, his thoughts, if taken to their logical conclusion, could make a reader imagine being replaced by the said system gradually. Moreover, what is stopping AI from creating multiple copies of us? AI could create dozens of me if it can duplicate me once.

Artisan, a company which bills itself as “Creating Highly Advanced Digital Workers … Using Cutting-Edge AI Technology,” is marketing itself as offering functional AI avatars replacement of salespeople, customer success, BDRs, recruiters, financiers, etc. with names like Ava, Liam, Noah, Olivia and others. Thought prospects are being bombarded by business development and sales now? Wait until a million Avas start getting in touch. In fact, among the features Artisan advertises per its artificial human replacements is “Sends 1000s of emails per month.” Buyers, decision-makers, V and C-levels take cover. Incoming! Pricing? From less than $1,000 (US) down to almost $100 per month. Do I get to talk to AI if I contact their sales department though?

Similarly, there is GoCust, which advertises itself as “Sales Teams Now Have Online Assistants.” It touts itself as a SFA, assistant, mapping and route optimizier, that gains cost and time advantages in order “to minimize the need for sales teams to constantly struggle with messaging, phone and email traffic… develop solutions to increase the rate at which customer conversations are recorded as data… to offer sales managers the opportunity to proactively manage their teams.” And these companies are not alone.

All of this has become possible, in the meantime, because storage and computation costs have fallen drastically and are basically cheap.

In the meantime, according to a study by Zendesk, which itself has jumped wholeheartedly into the AI space ($50/user/month), management believes AI is beneficial and helpful. “Four in five (81%) employee experience leaders now see AI as essential in boosting workers’ ability to tackle complex tasks,” reports Zendesk.

At this point, a forecast of five or six years for these technologies to be operationalized and become part of the mainstream seems frankly… conservative.

Things That Need To Go Away: Efficiency, Effectiveness And Productivity Ai Technology That Never Mentions The Potential For Creating Layoffs And Attrition. The Discussion Around These Technologies Needs To Be Brave And Candid.

The Amazon (AWS) AI Day

Articles Add Comments »

Jun 022024

Amazon, and its partner, Adastra offered a one day AI overview in Toronto recently. It was a hot topic that the presenters encapsulated well. Of course, they had an AWS bias and promoted Amazon’s services, but the context and implications were presented succinctly and clearly.

Below is a synopsis of the day’s presentations on Amazon’s Generative AI (Artificial Intelligence).

AWS (Amazon Web Services – the Cloud division of Amazon) Options:

1- Powered by Amazon’s Bedrock model

These are sitting on a host of services like Amazon Titan, Cohere, Meta’s Llama 3 and many more

AAS (as a service)/managed service and no overhead

2- Amazon SageMaker

Which is a ‘build your own’ platform

Amazon emphasises, however, that not one LLM fits all. It is important to assess your needs.

New Announcement: Amazon Q (an AI powered assistant)

Designed for business use cases | you can ask questions from it now

Use Cases:

CX/EX: onboarding, KYC, Routing, etc.
Personalization: Forecast/advisory, recommend
Text Analytics: Extract information from internal and external sources e.g. in the old days we had OCR with rules on top
Predictive Analytics: Extract data
Fraud Detection: identity fraud, anomalies

These were expensive to build and expensive to maintain and therefore fickle until now.

With AI all of the features are baked into the model and therefore a lot less development is required

Additional New Use Cases in Amazon Q:

Improve CX and EX
Increase knowledge of workers – think how important this is. We are all in exactly this business whether Marketing, Sales, Compliance or Analysis, etc.
Product Innovation and Process Automation including

- Data Extraction
- Natural Language Interfaces to Analytics
- Personalized Content Generation

Amazon Q (Suite of Gen AI Services) Portfolio – “AI Powered Assistant” General Availability: 30.04.2024 (new!) of 3 flavours:

Amazon Q in QuickSight (powered by Bedrock) can be considered a BI Service with AI Capabilities – structured data
Amazon For Business – also unstructured data
Amazon Q Developer – Assistance in writing code

Gen AI assistant for Accelerating software development and leveraging company data

These are being embedded in Amazon Services

Traditionally LLMs have been best for:

Broad World Information
Assisting Human Work (summarizing, teaching and Generalizing) as well as Offering knowledge, autonomous tasks and calculations

But Newly: Amazon Q in QuickSight:

Ask questions in natural languages and ML models interpret user’s questions and generate images and reports.
AI powered dashboard.
On-demand answers.
Can be extended to other Apps.
AI assisted Story telling (tells you what is going on and provides documents and slides to present the data)

These are available in different apps including your custom apps – Pricing (in USD):

Amazon Q Business Lite $3/user/month
Amazon Q Business Pro $20/user/month
Developer Free Tier
Developer Pro Tier: $19/user/month
etc.

Benefit: Ask yourself what is the productivity gain? We can see up to 75% productivity gain.

The AWS approach to Generative AI:

Enterprise Focus
Open Approach with both Proprietary and Open Source code
Cost And Power Optimized
Data Privacy and Security: Own Your Data/Access Control included/Permissions + Connectors to everything including Exchange/SFDC/Confluence/JIRA etc.
Based on the RAG Architecture

Generative AI and ML Considerations For Financial Services (because the day had a Financial Services bent nominally)

Approach? Is a human in the loop and is it possible given the amount of data? Is the input and output from structured or unstructured data?
Governance And Compliance?
Legal & Privacy: Think output validation and Reporting especially when using 3rd Party models
Monitor legal and regulatory scope consistently as the scene changes and proposals like SEC’s or European regulation become updated.

Things That Need To Go Away: The Only Name Tag For An Event To Be Missing Being Mine

Older Entries