Dr. Lisa AI

A.I. Solution Architect

CULTURE PARTNERS 

- I took the Revenue Team from zero use of AI to daily use of Generative AI for account and contact research, saving hundreds of hours a month, plus regular use of a Revenue Intelligence platform that uses AI for sentiment analysis, pattern recognition, deal and situation classification, and forecasting probability.
- I built the initial AI chatbot for the HR team to save them time from answering basic internal questions about HR policies and processes
- I showed the Products team how to make an AI chatbot that could answer questions about our products

Executive Vice President, Revenue Enablement & Technology

Executive Vice President of Products

Chief of Staff & VP Revenue Ops and Enablement

VP Sales Enablement

VantagePoint Performance

VantagePoint Performance uses Machine Learning and Data Science to inform cusromized sales training for B2B Enterprise Sales teams. Their research and methods are used by clients, consultants and other Sales Training companies.

Chief Executive Officer

EVP of Strategy & Infrastructure

VP of Customer Success

Altify

Altify automates key Revenue Enablement processes inside the Salesforce CRM, saving time and reducing stress for Salespeople, and improving information about how to win deals, for the whole company.

Principal Solution Consultant

Wurk

Wurk provides industry-specific Payroll and HR Management with an emphasis on leading edge tech like geofencing and automatic classification of activities based on employee movements.

Chief Operating Officer

ServiceMax

ServiceMax takes advantage of IoT (Internet Of Things) connected machinery by taking the data from that machinery and turning it into actionable insights and service requests that increase uptime and decrease repair and maintenance costs.

Senior Solutions Consultant, Japan

UKG (Ultimate Kronos Group)

UKG is a leading provider of Workforce Management solutions and insights.

Senior Manager, Business Value Realization, then Strategic Accounts Presales Consulting

Oracle

Specialized in demonstrating the architecture and dev tools to technical IT audiences

Senior Technical Product Consultant, PeopleSoft

Specialized in large, custom demonstrations for strategic accounts

Senior Sales Engineer, Siebel Systems

eInnovate

eInnovate was an early (pre-eBay) online marketplace built on the EJB Framework

Director Of Technology

Orica

Specialized in FI, CO and MM modules of SAP R/3

SAP Implementation Specialist

I help Sales and Marketing Leaders see their best uses cases for AI, and explain to their CIOs and Heads of Sales Enablement how it fits into their tech stacks.
I have several years of experience managing and building GTM (Sales and Marketing) tech stacks.

I'm an ex software developer who went into sales, then into management. I came back out of management to work directly with AI because the potential impact, positive and negative, was too big to ignore.

It seems like everyone is using LLMs as classifiers. 

At first glance, it appears to work well. It's certainly a massive step up from zero-shot classification with embeddings. 

However, as you begin to scale up the number classes (or have classes that are very similar) you are almost guaranteed to run into massive instability. Adding loads of few shots can help, but has its own caveats (the cost also becomes crazy). 

Traditionally, you would just train an encoder only model like BERT or RoBERTa - but this assumes (at least) thousands of examples per class. And given how emergent use cases can be nowadays it's just not super practical. 

Sentence Transformers Finetuning (SetFit) from HuggingFace is a really solid middle ground. The main goal is to fine-tune a sentence transformer embedding model to produce embeddings that are useful for a specific classification task. It has two main components:
1. A sentence transformer embedding model (the body)
2. A classifier (the head) - this can be as simple as a logistic regression model. 

To adapt the embeddings for a specific task, SetFit uses contrastive learning objective. Specifically, using pairs of sentences:
- Positive Pairs: Sentences in a pair are from the same class.
- Negative Pairs: Sentences in a pair are from different classes.

The model then updates its weights to pull positive pairs closer and push negative pairs apart in the embedding space.

The training objective makes SetFit extremely data-efficient. Sometimes you only need 8 examples per class!

It's also really fast for both training and inference (can run on a CPU just fine).

Eric Fraser

SetFit Overview

The way companies treat under-18s with their marketing tactics is just outright unethical. 

We researched this in our book and discovered that a consistent relationship has been identified between advertising exposure, materialism, and parent-child conflict. Not only are the negative effects of marketing products being seen but also how under-18s start to view both material possessions and themselves. 

Here are a few examples of where it is as Tom Tapper so eloquently puts it 'downright unethical':
- Using inadequacy marketing techniques (making people feel like they have a problem but then selling a product to say this will fix it) which when you are a teenager and pretty unsure of yourself is horrendous because you are making an already unstable person even more unstable and then telling them if they spend money, they can align with the trends and their friends.

- Sponsoring school events or placing OOH advertising near school grounds so in a place where they are often seen as protected, they are still being bombarded

- Placing colourfully branded products at children's height in supermarkets so they pick them up and then beg their parents to buy them for them

- Making collectible collections so make children want to buy more and more

- Using animated brand characters to attract children's attention such as Tony the Tiger

- Using influencers to market products. A detailed report requested by the European Parliament's Committee on Internal Market and Consumer Protection (IMCO) discovered that children due to their lower levels of advertising literacy are particularly vulnerable to the powers of influencer marketing because their perceived level of trust in peers and authenticity (https://lnkd.in/eCJ6PJxm). They often struggle to identify the marketing strategies adopted and their lack of education makes it harder for them to distinguish between branded and non-branded content. This in turn has led to four major concerns: influencing children's brand attachment and potential future purchasing habits, increased materialistic behaviour which results in financial difficulties in the future, exposure to content that is not age-appropriate and self-perceptions and expectations on what their bodies should look like. 

- The imagery used. For example, children being constantly shown images of women has led to an abundance of mental-health-related issues among young females who perceive this body shape as 'ideal' (https://lnkd.in/e6Ji3DYH)

It is time brands reconsidered their entire approach and put the correct governance parameters in place to prevent ruining our children's upbringing! 

Cracking piece of work Nice and Serious. Here is a link to the Global Action Plan petition - https://lnkd.in/epMkVre4 

#marketing #advertising #children

Megaproject alert: Abu Dhabi is building the world's first eSports island costing $1 billion!

The plan entails the construction of a GG Resort, a hotel with 200 gaming-equipped rooms, complete with a unique "parachute jump check in" system. 

Read more about the project below || https://zurl.co/MgSI #AbuDhabi #Entertainment

constructionweekonline.com • 3 min read

Abu Dhabi to build a $1bn eSports island with ‘parachute jump check-in’ hotel - Const...

Meta is hiring data scientists again. Here's what you need to know 👇

In 2024, they now assess these 4 areas in technical rounds:

📙 Programming - data manipulation with Python, R, or SQL
📗 Research Design - AB testing and/or causal inference
📘 Determining Goals and Success Metrics - KPIs and investigation
📕 Data Analysis - statistical modeling

To prep, use the following resources:
🔹 Prep for SQL: https://lnkd.in/gAmR4eQM
🔹 Practice AB testing cases: https://lnkd.in/g82dMJ77
🔹 Practice product sense and data analysis cases: https://lnkd.in/gzgB-dHT
🔹 Join the DS Interview Bootcamp led by me and Meta data scientist: https://lnkd.in/eiA5Ntdp

👉 Any other company guide you'd like me to cover? 👇

👉 Smash the 👍, share 💬, follow Daniel Lee for data and AI jobs!

datainterview.com • 3 min read

[2023] Meta Data Science Interview (+ Case Examples)

I'm amazed how quickly after a new AI methodology gets published the open source community builds code around it. Thursday Google shared the paper about Mixture-of-Depths about scaling in depth dimension while keeping the flops constant- similarly how Mixture-of-Experts does it in the width. The model can learn to route harder tokens through more layers, similar to how experts in MoE can specialize in certain domains. Saving the compute and memory on the model level.

Currently, there are at least four Mixture of Depths implementations from scratch and 2 upcycling to convert existing transformer models into MoD models. Just for the past two days. All are in 2 days of development but surprisingly most of them are working. The best is, all code and models are open source and you can build on top of them, and run them locally, offline on your own compute for the GPU poor.

Make sure you own your AI. AI in the cloud is not aligned with you; it's aligned with the company that owns it.


Big news, friends! Radious has gone national! And how cool that Forbes featured the announcement! Great writeup at: https://lnkd.in/gV87wa_i

tl;dr... for the last 8 months, we've been quietly piloting a "curation service" that opened up the opportunity for us to serve companies across the U.S.

And now, after months of iterating and fine-tuning our process under the radar, we're officially LIVE! We now offer unique, residential workspaces anywhere in the country!

So, how does it work?

1️⃣ Your company is planning an offsite, a brainstorm, or a coworking day with your team. Get in touch with us and share the deets (number of people, desired work amenities, etc.). Give us ~2 weeks of lead time.

2️⃣ Our curation team taps into our waitlist of properties (aka "secret spaces") and unlocks beautiful workspaces in your geo.

3️⃣ We present 3-5 options for you to choose from and bam! 💥 One-of-a-kind experience coming your team's way!

This curation service unlocks every geography across the nation, big or small. And when I say small, trust me. We have sourced killer workspaces in some VERY obscure places!

If you're interested in trying out a Radious workspace for your team, shoot me a DM!

Or visit our website at https://radious.pro to learn more.

New postdoc opening in my group! If you're interested in using cutting edge methods to understand how our social networks contribute to life outcomes through analyzing with one of the most amazing datasets in the world, this is the job for you!

Read more and apply here!

efzu.fa.em2.oraclecloud.com • 1 min read

Postdoc in Modeling Events in Connected Human Lives - DTU Compute

NEW: In an exclusive interview with Fortune, early OpenAI investor Vinod Khosla discussed his latest investment in Symbolica AI, founded by former
Tesla autopilot engineer George Morgan-which announced a $33M Series A today led by Khosla Ventures. 

Khosla said Symbolica offers a “very innovative approach” to one of the biggest challenges facing AI: How to develop smaller, more efficient models that can reason more like humans do. “We love people coming from left field,” he said.

Read the full story:

fortune.com

Vinod Khosla is betting on a former Tesla autopilot engineer who quit to build small AI models that can reason 

I heard my AI mentor Dr. Lisa Palmer say something brave and reassuring to a group of business leaders this week.

I'll paraphrase:

If your main motivation for learning AI is so you can lay off a bunch of your people, we shouldn't work together.

This was reassuring to me because I don't want to build a business that only grows by causing fear and grief. When you lay off a bunch of people, that's what happens.

AI will replace some tasks, and some whole jobs, like most new tech. But people don't have to be thrown away. People do things that AI cannot. Things that customers value. 

My AI Mentor Dr. Lisa Palmer sent me this tool that builds visual graphs of connections between research papers. If you spend as much time on ArXiv as I do, you will probably like this as much as I do. 
For example this is a graph it made in <1 second of papers which address "1.58 bit quantization of LLMs". (Something that has the potential to save a LOT of electricity and water while running today's massive AI models, btw).

www.connectedpapers.com


Ever wonder what an AI Startup CEO does with her days? 👇🏻

Just the big nuggets from this past week (sparing you the endless pile of admin tasks and legal/contracts work): 

➡️ To ensure that we create an unmatched value proposition for our clients, we’re finalizing a partnership with an amazing go-to-market startup - resourced with cutting-edge AI plus a massive network of seasoned fractional CMOs 🧠 - so that we can take organizations all the way from ideation to full market launch. 

➡️ I’m excited to be working on a LinkedIn Learning course on Generative AI for Board Directors! I’m finishing the draft script now and will be recording my sample for them next week 🎬.

➡️ To keep me energized and my creativity flowing, I spent ~30 hours in 2 days and built a card game for 11+ year olds called AI Adventures 🃏. It’s designed to teach key terms and spark the imagination through collaborative problem solving of 10 world missions (Save the Bees, Build Eco-Friendly City, Health Hero, and more!) Building this reminded me of how much positive potential exists for the next generation if we teach them about human plus AI partnering! ✨

➡️ Amazing hands-on work! 

Onsite with an energy company who is incredibly well setup to explode business value through AI. I left there feeling genuinely excited for them and their future! ⚡

Public relations and criminal investigations problem solving with a progressive and responsibly minded law enforcement agency. The way that they are thinking about serving their community makes me feel privileged to serve them! 🛡️

Building out a custom GPT for a technology association client who is looking for best practices across piles of similar but different policies. LLMs are perfectly suited to this task! 🤖

And more! Insurance, professional services, marketing, law office, banking, SaaS provider, etc…the list is long and varied! So much value being created with AI 🤯.

➡️ Editing a critical product of ours that has passed MVP1 phase and we’re applying the feedback learnings into MVP2. This is CRITICAL to a major pitch we have upcoming so seeing this come to life was exhilarating 💪🎉.

➡️ Pitch building and iterating. Creating this document has been a powerful focal exercise to clarify exactly who and how we want to serve. 📈

🌟🤝 Highlight: Meeting my fantastic CTO, Eric Fraser, in real life for the first time after knowing one another virtually for years. He’s brilliant, great with customers, and a truly wonderful human. I’m thrilled to have him on my team!

Please don’t add up the hours in the week. It’s A LOT. And I’m currently writing this from a plane on the way to San Diego for a keynote. 

This life is all consuming. It’s exhausting. It’s also wonderful!

And I can only do it by surrounding myself with amazing colleagues and partners - and the endless love and support of my family and friends. 

Love and light to all of you! 
Happy weekend ☀️

#entrepreneurlife #lifelonglearning #makingadentintheuniverse

What are people saying about the 2024 SIM RTP Tech Connect keynote speaker, Dr. Lisa Palmer?

"One of the best presentations I've ever heard on AI was from Dr. Lisa Palmer, who was the first person to give me a REAL answer on 'How can we address biases and gaps in our AI data."

"Love her enthusiasm and energy on the subject."

"She makes AI easy to understand."

"Wow!... Great Speaker and she truly knows her subject... I could have listened to her for hours..."

Dr. Lisa has worked for top companies including Microsoft, Splunk, Gartner, and ConocoPhillips, with a latticed career journey that includes: 

🌀 Engineer and IT practitioner
🌀 Technology sales leadership
🌀 Technology researcher and advisor
🌀 Industry consultant
🌀 Executive roles as both CIO and CMO

✨ Don't miss out on an exciting agenda and informative presentations! ✨

Register today and secure your spot at the 2024 SIM RTP Tech Connect Conference happening on Thursday, April 25 from 11:30 AM to 8PM!
https://lnkd.in/dxwESH3

➕ Attendance is complimentary for SIM Members and IT Practitioners. Your ticket includes conference access, breaks, optional lunch, social and dinner. 
➕ Non-Member IT Practitioners: $99 ($149 after March 29)
➕ Vendors interested in attending should contact Matt Minoski, SIM RTP's Philanthropy Chair, for sponsor information that includes conference registrations.

#SIMRTPTechConnect #SIMRTPEvents #ITPractitionerEvents #BiasinData #FutureofTech #DataDrivenWorld

Biological Intelligence compared to AI: Our dog is almost 14 years old, so to keep his mind sharp we put his treats into a "dog puzzle" that makes him do simple logic to get the treats. 
(This is what Asian parents do).
There are 3 levels of puzzle, he's mastered level 1 so we bought him level 2. He was starting to master level 2 so my wife took some pieces from level 1 and added them to level 2 to make it harder.
In response our dog took all the level 1 pieces and tried to hide them from us 😆 

I'm going to photograph the puzzle from all angles, upload the pics and a schematic of the puzzle to ChatGPT4, and see if it can provide instructions on how to get the treats inside the puzzle.

The trend for Tech Giants now is to add "Big Name" AI talent to their leadership teams and boards. Andrew Ng is definitely a Big Name in AI.

It would be interesting to see how someone like Andrew Ng actually affects decision making about AI at a company like Amazon, through being on its Board. I hope someone interviews him about that about 6 months from now.

reuters.com • 2 min read

Amazon adds AI expert Andrew Ng to board as GenAI race heats up

Simply put, easy to remember, and worth remembering.
(From Patrick Maroney )

Yes, researchers are trying to make AI learn through sensors, like human babies learn with their 5 senses, but that's not here yet. Our AI needs tons of data to be useful.

At RTM 2024 CIO event in San Diego and among many topics, we are diving into AI.  While almost every sector will be impacted, my mind is focused on how critical it will be to continue enhancing and expanding our digital literacy skills and how important ongoing conversations around ethical use and opportunity will be.  It must also be a broadly accessible discussion, especially as I consider our rural communities.  Feeling extra thankful for my liberal arts education today - excitement, optimism and caution, words being shared by many of my colleagues.  

Sora - sharing research of AI-generated content with just a prompt.  Shared by Dr. Lisa Palmer, our opening keynote.

Comedians making fun of something is often a good indicator for how the general population is starting to feel. AI executives and advocates should probably be careful with over-promising.

Thanks Tim Sullivan for sending me this clip, below.

Human experience of AI and honesty about its current limits are good things to remember when you work in AI.

One of my personal, favorite tests before I say something about AI to a client is "will this demystify it, or make it sound more mysterious and complex?" 

It's not helpful to talk about it as if it's some kind of God or Superbeing. It's also not helpful to be totally uncaring about job losses.

https://lnkd.in/g5iNDtRb

 Working in AI: It's taken me about 
- 4 weeks of "thinking and pondering time" 
to see a possible approach that will take me 
- 4 hours to build a proof of concept for. 
If it works it can save 
- 1000s of hours. 

I honestly don't know if this marks me as slow or fast. Does anyone else experience this pattern working in AI? Specifically the multiple days thinking and pondering about how to do something (while also doing a bunch of other tasks).


Quantum Computing breakthrough by Microsoft and Quantinuum. A big one. 

We might be 5 years or less away from Quantum computers being affordable as business machines the same way that regular computers are today.

This breakthrough addresses the reliability of qubits, which are hard to control due to their quantum nature. 

The other hurdle is how to make qubits that are "boring and affordable", instead of requiring highly exotic materials and/or highly unnatural conditions that don't exist on Earth outside of laboratories.

Quantum computers could dramatically reduce the electricity needed by AI, while making it more affordable and reliable.

https://lnkd.in/edVd84t4

Microsoft が、アメリカ、韓国、インドの選挙に対して、中国がAI生成コンテンツを使って妨害しようと計画しているというレポートは発表。

すでに台湾総統選挙で、AI生成コンテンツによる選挙妨害があったとのこと。

https://lnkd.in/gyymwHEx

どこが仕掛けてくるかは別として、社会気運の醸成に、AI生成コンテンツ＆SNSを使った手法は、今後スタンダードになるでしょうね。

インフルエンサーマーケティングが、生成AIマーケティングになると思えば、分かりやすい。

当然、政治的な目的以外にも、商業的な目的で同様の手法は取られ、ヒトの思考や嗜好を操作されていくことにもなる。

では、対抗策は？

Taplio

Eric Fraser's Linkedin Analytics

Eric Fraser

What is Eric talking about?

Eric Fraser's Best Posts (last 30 days)

Want to drive more opportunities from LinkedIn?

Taplio

.css-1jgreyo{background-image:linear-gradient(284deg, rgba(39,43,72,1) 0%, rgba(33,47,183,1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Eric Fraser's Linkedin Analytics

Eric Fraser

What is Eric talking about?

Eric Fraser's Best Posts .css-qqfgvy{font-size:var(--chakra-fontSizes-md);}(last 30 days)

Want to drive more opportunities from LinkedIn?

Eric Fraser's Linkedin Analytics

Eric Fraser's Best Posts (last 30 days)