We’ve all seen firsthand how technological advancements can drive profound changes in our economy. In this era of digital transformation, one of the most promising developments in artificial intelligence (AI) is the rise of foundational models, like GPT-4 from OpenAI. These large-scale machine learning models, trained on diverse Internet text, show an unprecedented versatility and are now being considered as foundational building blocks for a range of AI applications. In this blog post I will present 5 reasons why foundational models represent the future of AI.
Here is a quick finger counting hand pose estimation model. Not perfect, but cool…
We started on a project tonight to build a computer vision model that will classify a few family members including the dogs. We used Teachable Machine to get a model built, and will now be exporting a Keras model to run in TensorFlow. Love teaching the kids the power of AI. More to come…
I believe in on boarding ways of thinking via models to help drive faster, hopefully consistent practical decisions…to quickly say, ah it’s just another one of those. Most models on their own will lead you astray. However, applying multi-model thinking has statically improved outcomes. Here’s another model to add, from Shane Parrish’s The Great Mental Models Volume 2.
The benefit of the doubt may never be equally distributed; however, everyday individual choices are.
Force Field Analysis essentially recognizes that in any situation where change is desired, successful management of that change requires applied *inversion. Here is a brief explanation of this process:
1) Identify the problem
2) Define your objective
3) Identify the forces that support change towards your objective
4) Identify the forces that impede change towards the objective
5) Strategize a solution! This may involve both augmenting or adding to the forces in step 3, and reducing or eliminating the forces in step 4.
12 Places To Intervene In A System, To Drive Systemic ChamgeAfter reading quite a few books on systemic racism. I was compelled to find a book on discipline of Systems Thinking. I found "Thinking in Systems" by Donella Meadows to be a highly read and rated choice on the topic. Given the complex nature of systemic racism and racist actions, how do you tackle it. I believe systems thinking can provide a framework for doing just that. Given that we can’t just change a system directly, in "Thinking in Systems," Donella Meadows outlined a list of interventions you can lever to influence the system. She sorts the leverage points in increasing order of effectiveness, from the easiest to lever/least long term impact on the system; to the hardest to lever/most effective to long term impact on the system. The easiest and least effective is effecting Numbers (e.g. effecting #'s and %'s); the hardest and most effective is Transcending Paradigms (which is almost spiritual), but 2nd to the hardest/most impactful is Paradigms (i.e. changing the societal culture around how we consider each other). I think racism needs to be attacked from the top and bottom, that is starting with Numbers AND Paradigms; converging where they do.
Read Time:33 Second
Fun little project his weekend, building a object detection and classification solution for less than $100. Though this pic only shows “person” and “book” classifications, the model can classify some 90 objects! The Tensorflow Lite model is running on a 4GB Raspberry Pi 4 w/ 128GB Sdcard. The camera is a Arducam, which I need to work on the resolution for but it didn’t impact the detection or classification, and ran at ~2.0 fps. Running on a Pi I have a give and take between model performance and accuracy, given the limited resources, but will push to see how resource hungry a model I can run on it. More to come…
I enjoyed doing this segment with Jeff Lahens, my former business partner, and Jarvis Green, good friend and 2X Super Bowl champion with The New England Patriots… now owner of Oceans 97.
Read Time:0 Second
This."No muscles without strength, friendship without trust, opinion without consequence, change without aesthetics, age without values, life without effort, water without thirst, food without nourishment, love without sacrifice, power without fairness, facts without rigor, statistics without logic, mathematics without proof, teaching without experience, politeness without warmth, values without embodiment, degrees without erudition, militarism without fortitude, progress without civilization, friendship without investment, virtue without risk, probability without ergodicity, wealth without exposure, complication without depth, fluency without content, decision without asymmetry, science without skepticism, religion without tolerance, and, most of all: nothing without skin in the game." ~ Nassim Nicholas Taleb
When seeking to develop an innovation strategy, here are some questions you should get answered.How do I see emerging trends before they become problematic?How do I generate a robust pipeline of new growth ideas to consider?How do I identify and focus on the highest-potential opportunities in areas like blockchain and AI?How can I motivate traditional company management to realize the need for digital transformation?How do I evaluate competitive signals in a noisy, buzzword-filled market?How do I get the middle layer of my company to embrace change?How do I bring outside ideas into my organization?When does it make sense to be a fast-follower? And when does it not?How do I decide whether to build, buy or partner?Should I start a venture capital fund?
IntroductionHypeLimitationsBiasAdversarial attacksImpact on developing economics and jobsA realistic viewGoldilocks rule for AI:Too optimistic: Sentient/AGI, killer robotsToo pessimistic: AI cannot do everything, so an AI winter is comingas opposed to the past, AI is creating value today.Just right: Can't do everything, but will transform industriesLimitations of AIperformance limitations. (limited data issues)Explainability is hard (instructible)Biased AI through biased dataAdversarial attacksDiscrimination/Bias BiasesBias against women and minorities in hiringBias against dark skinned peoplebanks offering hiring interest rates to minoritiesreinforcing unhealthy stereotypesTechnical solutions"Zero out" the bias in wordsUse more inclusive dataMore transparency and auditing processesMore Diverse workforceAdversarial attacksMinor perturbation to pixels can lead and AI to have a different B output.Adversarial defensesDefenses exist; incur some performance costThere are some applications that will remain in an arms race.Adverse uses of AIDeepFakes, fakes can move faster than the truth can catch upUndermining of democracy and privacy, oppressive surveillanceGenerating fake commentsspam vs. anti-spam, fraud vs. anti fraudAI and developing economiesAI will eliminate lower rung opportunities. The development of leapfrog opportunities will be required. Think how countries jumped to mobile phones, mobile payments, online education, etc.US and china leading, but still a very immature space.Use AI to strengthen country's vertical industries.More public-private partnershipsinvest in educationAI and JobsAI is automation on steroids.SolutionsConditional basic income: provide a safety net but incentivize learningLifelong learning societyPolitical solutionsConclusionWhat is AI?Building AI projectsBuilding AI in your companyAI and society
IntroductionStarting an AI projectWorkflow of projectsSelecting AI projectsOrganizing data and team for the projectsWorkflow of a machine learning projectHow do you build, say a speech recognition engineKey Steps:Collect Data: people saying "Alexa", and other wordsTrain model: learns A to B mapping... audio clip to "word"many iterationsDeploy the model: implement in to a smart speakerWill collect new data (get data back), to maintain /update the modelHow do you build, say a self driving carKey steps:Collect Data: images - > positions of other cars, draw rectangles around carsTrain model: need to iterate and precisely identify carsDeploy model: may learn that golf carts are identified and positions well. keep iterating.Workflow of a data science projectoutput: actionable insightsOptimize a sales funnelKey steps:Collect Data: where are people coming from, time of day, machines type, etsc...Analyze the data: Iterate many time to get good insights insights from the data collected.Suggest hypotheses/actions: Deploy changes, re-analyze new data periodically.Optimizing a manufacturing lineKey steps:Collect Data: clay supplier, mixing time, ingredients, lead times, relative humidity, temperature, kiln duration, etc...Analyze the data: Iterate many time to get good insights insights from the data collected. Suggest hypotheses/actions: Deploy changes, re-analyze new data periodically.Every job function needs to learn how to use dataUse data to optimize workflows through data science based analysis, and to take on tasks with machine learning (remember less than a second), Inputs (A) to Output (B).From Sales, recruiting, marketing, to agriculture, and beyond DS and ML are having huge impactsHow to choose an AI projectBring together a cross-functional team knowledgeable in AI, plus domain experts.Brainstorming framework:Think about automating "tasks," vs automating "jobs."what are the main drivers of business values?What are the main pain point in your business?Note: you can make progress without big dataHaving more data almost never hurtsData makes some business [Google, Facebook, Netflix, Amazon] defensible.With small datasets, you can still make progress. The amount of data you need is problem dependant.Due diligence on projectWhat AI can do + Valuable for your businessTechnical diligenceCan AI system meet desired performance (e.g. accuracy, speed, etc)How much data is need to meet performance goalsEngineering timlineBusiness diligenceCurrent business: Lower costsCurrent business: Increase revenue ( getting more people to check out)New business: New product or business*Ethical diligence*money vs impact on societyBuild vs. buyML projects can be in-house or outsourcedDS projects are more commonly in-houseSome things will be industry standard, avoid building those."Don't sprint in front of a train."some times it makes sense to adopt another's platform or approach than to build your own. resource constraints, capability constraints...Working with an AI teamSpecify your acceptance criteriaGoal: defects with 95% accuracy...How do you measure accuracyTest Set (n1000): labelled training dataset to measure performance. Training Set: Pictures with labelsLearn mapping from A to BTest Set: Another data set to test the mappings. Often more than 1 test set will be requested.Pitfall of expecting 100% accuracy. Discuss with AI engineers what's reasonable.Limitations of MLInsufficient dataMislabeled dataAmbiguous labelsTechnical Tools for AI teamsCPU vs. GPU [Great for deep Learning/Neural Networks] NvidiaCloud vs. On-prem, ....Edge [Processor, where data is collected.]
Retailers are feeling significant pressure as digital takes a greater and greater hold of the industry. Many say that digitization actually brings demonetization. This will result in a massive shift in the share pie for retailers. What once was a great traditional retail business, will become a much smaller primarily online business. I thought I'd take a stab at visualizing that. Thoughts?
1. Classification (probability estimation or scoring): binary or categorical.Attempt to predict, for each individual in the population, which of a (small) set of classes this individual belongs to. Classification will bucket individuals, and scoring will provide quantification of likelihood of being in a particular bucket.