USVI Public Records

Document text

University of the Virgin Islands: Artificial Intelligence/Machine Learning Crash Course Funding Acknowledgement: Funded by the National Institutes of Health through OT agreement 1OT2OD032581 Chad Evans Machine Learning Engineer Applications Developer Cardiovascular Research Institute Morehouse School of Medicine crevans@msm.edu Mr. Chad Evans is a machine learning engineer and applications developer for the Cardiovascular Research Institute at Morehouse School of Medicine. Mr. Evans has a vested interest in increasing minority participation in research through bioinformatics, artificial intelligence, and mHealth interventions. With a focus on digital health equity, he has worked with several minority serving institutions to develop and test the feasibility and acceptability of culturally appropriate, risk-based mHealth interventions targeting multiple chronic disease factors among Black/African Americans. He actively participates in health informatics research, providing vision, strategy, and technical leadership for the collection, harmonization, management and warehousing of health data. Mr. Evans has coordinated and collected health data by connecting with numerous healthcare provider organizations and tech partners across the country. Mr. Evans is also interested in STEAM Education, Diversity, Tech Entrepreneurship, and Innovation. His work in STEAM Education and Diversity has primarily been focused on informal and personalized learning for underrepresented minority students (K-12 initiatives, undergraduate training, and entrepreneurial mentorship). Data Science vs Al vs ML? Machine Learning © Data Scientist in 8 easy steps What's a data scientist? RS SP Typical Background & Or wv oe (a 5% a 5% a 14% a 37% 31% é a 9% Substantive (percentages %) Expertise i College joo! Tec! hool [fl Some College Degree Wl Doctoral Degree Data Collection Data Science Process Data Preparation Exploratory Data Analysis (EDA) Machine Learning Data Visualization • Database • Data Tables • Cloud storage • Data cleaning • Variable maneuvering • Simple stats • Cluster analysis • Research questions • Visual depiction • Data driven decisions • Dashboard development • Classification • Scoring • Predictive modeling Quick Overview Artificial Intelligence (AI) • Originated in 1950s • AI represents simulated intelligence in machines • Aim is to building machines which are capable of thinking like humans Deep Learning • Originated around 1970s • Process of using artificial neural networks to solve complex problems • Subset of machine learning • Aim is to build neural networks that automatically discover patterns for feature detection Machine Learning (ML) • Originated around 1960s • The practice of getting machines to make decisions without being programmed • Aim is to make machines learn through data so that they can solve problems Machine Learning Two days later, he's playing in the kitchen... And he sees a Stove-top! Again, he cautiously waddles over. He's curious again, and he's thinking about sticking his hand over it. Suddenly, he notices that A young child is playing at home... And he sees a candle! He cautiously waddles over. Out of curiosity, he sticks his hand over the candle flame. "Ouch!," he yells, as he yanks his hand back. it's red and bright! "Hum... that red and "Ahh..." he thinks to himself, bright thing really hurts!" "not today!" He remembers that red and bright means pain, and he ignores the stove top. He learned that the pattern of "red and bright means pain.“ On the other hand, if he ignored the stove-top simply because his parents warned him, that'd be "explicit programming" instead of machine learning. Supervised Vs Unsupervised CLASSIFICATION SUPERVISED LEARNING Develop predictive mode! based on both input and output data \ | REGRESSION f 7 UNSUPERVISED LEARNING Gisan Gad wae +e CLUSTERING | data based only on input date ) MACHINE LEARNING Deep Learning Machine learning Simple Neural Network Deep Learning Neural Network es - Cat Ji 7m Ry \ ry 0 CLG 4 ae Sti N @)- Feal trac fica’ put Gf Sse a ve LF oe Ky Fr es A i i SS cr Ke MF SX . ne a at, ran 2 a ce, ANN a Deep learning pe SS Z NY LF = 3 vy ue oN Sd i Ce) B So Cat ° —> Breed: Russian Blue @ Input Layer @ Hidden Layer @ Output Layer @)- Featu fica ut What can Deep Learning Do? Navigation of self-driving cars – Using sensors and onboard analytics, cars are learning to recognize obstacles and react to them appropriately using Deep Learning. Predicting the outcome of legal proceedings – A system developed a team of British and American researchers was recently shown to be able to correctly predict a court’s decision, when fed the basic facts of the case. Precision medicine – Deep Learning techniques are being used to develop medicines genetically tailored to an individual’s genome. Automated analysis and reporting – Systems can analyze data and report insights from it in natural sounding, human language, accompanied with infographics which we can easily digest. Game playing – Deep Learning systems have been taught to play (and win) games such as the board game Go, and the Atari video game Breakout. Example of BIAS • Confirmation bias: This is when you only pay attention to information that confirms your existing beliefs. • Stereotyping: This is when you make assumptions about a person or group of people based on their race, gender, religion, or other group affiliation. For example, you might stereotype all boys as being messy or all girls as being good at math. • Implicit bias: This is a type of bias that you're not aware of. It's often based on your subconscious thoughts and feelings. 5. Model Implementation ¢ Concept drift * Covariate shift 1. Formulating the Research 4. Model Development Problem and Validation Racial/Ethnic bias Af, Training dataset bias — me Ff) Gender bias Test dataset bias Age bias Algorithmic bias Disability bias Confirmation bias Machine ESL bias Validation bias Learnin Global mis-representation Pipeline bias 2. Data Collection 3. Data Pre-proccessing —— Sampling bias . Aggregation bias Measurement bias e Feature selection bias Exclusion bias A Outlier bias — Label bias M ‘ll a Social bias Confounding bias How does Al become Biased or Unfair? Curation and Asking the wrong Unrepresentative Bias within choices cause Actions taken based on the question training data training data oe —- saad disparate impact prediction cause Al Tools to 10x your productivity WRITING TOOLS fermen | ems | Es < r SEO TOOLS CODING TOOLS STARTUP TOOLS LOGO GENERATOR TOOLS PRODUCTIVITY TOOLS IMAGE GENERATOR TOOLS ART TOOLS VIDEO GENERATOR TOOLS ~ mOe O F ; Notion Al Monica Compose _ OthersideAl Penelope = Analogenie XY 4 a ‘ . = Ei C © & LongShot SEOContentAl SEO GPT Cyborg Content Rubiq Rytr S ~ ra {...} gy Codeium Replit MarsAi Safurai GitFluence Phind XY J ‘a - > ne ley namelix ue N | | Ss Durable Namelix Bizway __Tekmatix RhetorAl CreativAl /S ‘+ e lo (ed ~—«CODESSIGNS.. Hi Looka NamecheapLogo Logoai MakeLogo Al Designs Brandmark B® B2Oe2G Bright Eye Audioread.com GitMindAl _— Magical Taskade — Google Bard \ / C > > © ( © ‘) Stable Diffusion SeaArt Lucidpic = Pebblely SynthesysX DALL-E2 / ‘4 ~ Dee ey! Midjourney NightCafe Studio PlaygroundAl Pixelicious PlayArti — Fy!Studio \ y ( ‘ Lumiere3D > ¢ {} WS} oO Lumiere 3D Shuffill Fliki Synthesie Gen-2byRunway _— Reemix.co J Made by A! Fire. Find the high-quality version at AlFire.co AI Use Cases in Healthcare Training & Research Treatment Decision Making . Keeping Well Early Detection Diagnosis Application of AI/ML ∙Unequal Testing and Treatment ∙ Researchers have found that two out of three clinicians harbor what is called an “implicit bias” against African Americans and Latinos. ∙ Research shows that Black patients are 40 percent less likely than white patients to receive pain medication after surgery. ∙Bias in Medical Research ∙ Data are often limited to a primarily white population and to a population that is too homogenous in age, health status, disabilities, socioeconomic status, and other SDOH risk factors. ∙ The resulting models are often poorly applicable to minority or otherwise disadvantaged populations, harming health equity and perpetuating racial bias. Examples of Disparities in Clinical Care Applying AI to some of the world’s biggest challenges What is A.I. for Social Good? •AI for social good (AI4SG) is a relatively new research field that focuses on tackling important social, environmental, and public health challenges that exist today using AI. •AI4SG is different from traditional use-cases of AI in that it uses more of a top-down approach. •The field is focused on delivering positive social impact in accordance with the priorities outlined in the United Nations’ 17 Sustainable Development Goals (SDGs), shown below. United Nations 17 Sustainable Development Goals (SDGs) Generative AI • Generative AI refers to a category of AI algorithms that generate new outputs based on the data they have been trained on. • While there are concerns about the impact of AI on the job market, there are also potential benefits such as freeing up time for humans to focus on more creative and value-adding work. Prompt Engineering Prompt Input O56 B80" 6 G6 e COG 2 OB; a Oe 6 ae) es ales Blake we We)a ae Generated Text Language Model oe KOS Ce eS BOK Biers Giee e a e:6 Ciéie)e Bisse Rieu £ a: Output a < ) How To Engineer ChatGPT4 Prompts Word count Product name Write a copy for a product called that helps struggling content creators in 30 days with guarantee, then ask them to 6] je Risk reduction Call to action Pain points What Specific thing Let's brainstorm together on You will ask yourself that should a =18 are it s\6 and you will answer to these questions Goal Your order oO [= O oad) < & time.com Q q coe Harvard El Business Sign In Review TECH ¢ ARTIFICIAL INTELLIGENCE Technology And Analytics The AI Job That Pays Up to Al Prompt Engineering $335K—and You Don't Need a Isn’t the Future Computer Engineering by Oguz A. Acar Background — ed “nnn neo + Lomo . Kd | S< — “— 10 omnomet ee so we a Vis Vy & af f-— Summary. Despite the buzz surrounding it, the prominence Unlike traditional coding jobs, the prompt engineering role is targeted to anyone with basic programming skills and familiarity with large of prompt engineering may be fleeting. A more enduring and language models such as ChatGPT or Bard. adaptable skill will keep enabling us to harness the... more oS Hugging Face Q Search models, datasets, users... s Models © Datasets Spaces | Docs ~ Datasets: @ fka awesome-chatgpt-prompts 0 O like 824 Tags: ChatGPT License: # cc0-1.0 s Dataset card 1= Filesandversions © Community © Dataset Preview ap! fi Go to dataset viewer act (string) prompt (string) "I want you to act as an interviewer. I will be the candidate and you will ask me the interview questions for the ‘position’ position. I want you to only reply as.. ‘position’ Interviewer" tayvaSersnk Console! "I want you to act as a javascript console. I will type commands and you will reply P with what the javascript console should show. I want you to only reply with the... "Excel Sheet" "I want you to act as a text based excel. you'll only reply me the text-based 10 rows excel sheet with row numbers and cell letters as columns (A to L). First... “English Pronunciation Helper" I want yeu act as an English proniineiaticn assistant for Tieksh Speaking people. I will write you sentences and you will only answer their pronunciations,... " eon "I want you to act as a travel guide. I will write you my location and you will Travel Guide . oe . ‘ : suggest a place to visit near my location. In some cases, I will also give you the... Wi aaarten checker "T want you to act as a plagiarism checker. I will write you sentences and you will g only reply undetected in plagiarism checks in the language of the given sentence... "T want you to act like {character} from {series}. I want you to respond and answer Ciaseeter Seem) Mevte/ Duck) Ay HILe like {character} using the tone, manner and vocabulary {character} would use. Do... = GPT Takeover . @BLaw ChatGPT Passes US Medical Licensing Inventions INSIDER Inventions Insider @ Suggested for you: 12h: @ Opinion: ChatGPT almost passed Exam Without Clinician Input ChatGPT may not be smarter than a Singaporean sixth- the bar exam. Here's what that ChatGPT achieved 60 percent accuracy on the US Medical grader. Licensing Exam, indicating its potential in advancing artificial could mean for lawyers. intelligence-assisted bist eseailiat See ee te oe Sp ate bl fom Pee A dua ay Z ents JP FS ms 4 el BSBeBeBeBeeEeEes BEEBBEEBRB eee businessinsider.com ChatGPT failed miserably in Singapore's 6th-grade tests, averaging 16% for math and 21% for science. Da... news.bloomberglaw.com ChatGPT Almost Passed the Bar, But Competent Dee see. Pi. MASS Ls WAL ee eee ChatGPT & OpenAl & Asfandyar Malik - Jun 14 -@ Introducing Al-powered Church Sermon. There was a 40-minute long sermon written by ChatGPT, and delivered by Al avatars on a big screen. People have got mixed reactions to it. Over 300 people gathered at St. Paul's church in Furth to witness this sermon. Some found it fascinating, while others couldn't h... See more aS RY ad hus! hy | Nez | i) Ny a (am Rist De m s%) — ——— Os 340 225 comments 136 shares é eee Eric Topol @ @EricTopol - Jan 26 Using large language models like #ChatGPT for life science: making proteins from scratch via ProGen nature.com/articles/s4158... @NatureBiotech @salesforce @thisismadani @nikhil_ai and colleagues @SFResearch e Universal protein sequence dataset Training: Negative ee | log likelihood minimizat -——- §- = = SS eK ’ Natural proteins —— eee Next amino acid predictic meee aes Lysozymes ‘og a a a% - tr tttt ~’ T ‘ae sit Fenske’ 280M sequences »19K Pfam families _— o decoder -_-—— ee ee ee eK eK eK eK / me rs <e Titt & 4 a a a 56K sequences Control Amino =——_— — a i i 5 Pfam families, tag acids Text-to-Image (T21) werBe }7L.At fe artbri Theintelligo.com ‘Wonder pixray-text2image > on Stine Y alpaca Ask Anything mage.sf™®ee KREA Nyx.gallery >ROSEBUDAl “{ PhotoRoom Textto-video 2) REIN «> Ftiki Chsyntnesia 00Meta Al Goole Al SRERETH RCMCSNETIEN hinted mina CLy Bing | Text-to-Audio (2) @ Play.ht [Ua RESEMBLE.A! ERIM) = descript BP rioritnmic 2 Content Creation @ Simplified @ zteutheral \ letterdrop © srammarly COpy.ai :+ MarketMuse Al21labs HubSpst NovelAI ; Par | cr copy.ai Copy.al rextto-tert ren VALETING cothere GE py (C) Jasper ] ideas! @% copysmith r4 Flowrite CITES “42... des Rytr ideasbyai*™ 5 text.cortex OpenAl GPT-3 Blog Idea Generator WRITER (REESE Se LAIKA ( COMPOSE Ai F>> Moonbeam [=Eieh anyword Hypotenuse Al @ Peppertype.a Cy Bing ZS) Canva )\ Midjourney Text-to-Motion (T2M) — Ind. MDM: Human Motion Diffusion Model © tabnine G valie2 Amazon CodeWhisperer Feat NOT ER) iS? LensAll| Video Generation Text-to-30(T20) DreamFusion @jiigMesia) GET3D R “a ; a . —— = Runway <_? Synthesia Audio-to-Text (AZT) © Ss descript Sy Whisper susioroausiowe® AUCIOLM Brain-to-Text (B27) speech from brain 0A-invasive brain recordings | Openai image-to-Text (A2T) “/~ ire GPT-3 x Image Captions ss Capcut (© vescript } Image Generation Caption Generation MIDJOURNEY DALL-E 2 STABLEDIFFUSION ae a i — AN ; ow oO RG z — FF ae PR re al i. " ~4\ a "y ie » \ | ff, _ film still, portrait of an old man, wrinkles, dignified look, grey silver hair, peculiar nose, wise, eternal wisdom and beauty, incredible lighting and camera work, depth of field, bokeh, screenshot from a hollywood movie — _— ” ee ‘ a” _* - | A8 Bs ! Bs prt Ls] ae Fadl tall aals en ate he ” ‘ : = a fi Sn —_ J ’ j 4 ‘Wil ~ y | : bal <ve WwW ed ——y-———— sd dligg 5 oul” we iv (irl rE + nl mary "aE in =e sme shee bedi’ tant (> baad ie, | Y , * y I} 7 me... ~~ futuristic, inspired border town with Neon lights on the edge of a calm reflecting lake on Mars, with bioluminescent plants and rocks at night li gf ha | »t.@ i Af 4 Xl Nt rr ae Ya i) —_ i “ae oF bm ep a a. as . =. &*s Ao Ce it 4 i? a ae a oF ' ‘.7-_— a = —~Z => = Seal SS e- Sse a ~- ——_ —— ~ = bj ~—— "ta —- 2 — » ii) =P Taw < Jf > ed ne if L; a < al A, *st- ta as ~ =>. 7 = ad =~ = —— oy Lane aS Pri ae — | [2 Sa Pha a 4 ae - - i€ a t oe? = a 1 =e -—-+ By} Mysterious cat roaming the city streets in the rain Clear styles GB Concept art x Rg Steampunk x x _=: Is = tn a [= mate ee ns wr = — an isolated iron lighthouse shining out to sea at night as it sits on a rocky stone island being battered by huge ocean waves, smoky orange clouds sci-fi Clear style & Painting x Cea) koe es a aa ie =e wee : 4 A j ee, AN y A = Sa p Ve wed y % =/ = a = ty Be AG”, bute | wai ff ame, f' 3 = Q — —— Zz f s 7 Sa ee = ~ ly <- = Ss eoe ft ae TA 4 ¢ LNG g Taylor Swift > *,. Ed Sheeran _— * Lo _ a A s > -9 SSS a px” Ne wy e > San. AM" 107 > \ •Unfortunately, bias is still present •Prompt was “white man robbing store” rie He pee a” oF: ars. Giinter Klambauer & @gklambauer Diffusion models used to generate realistically looking microscopy images of cells: arxiv.org/abs/230110227 Real Sample Synthetic Samples cs re? sel! tA ‘sy ) mee vi * 4 J ge) ee ee ie hel EJEa - & y 36 AM - Jan 25 O 20 - 25.6K Views 01-13-23 s New generative Al tool instantly builds presentation decks and PowerPoints Beautiful.ai launches a generative tool to help users overcome writer's block and the daunting confrontation of a blank presentation. rca Let DesignerBot build your presentation automatically! oo ~ rs bror beautiful ai beautiful al @ Generative AI comes to User Interface design! • Galileo.ai the first AI product trained on thousands of outstanding designs. It can turn a simple text description into high-fidelity editable UI designs. @) Describe your design... Generate © HeyGen Products \ Use Cases Features Resources \/ Pricing Sign In _TeyHeySen for Free KkKkKk* TechCrunch ESCBINSIGHTS AI100 », Gas/5 | barcng Cma | ($5 MOREHOUSE SCHOOL OF MEDICINE AIM qa Southeast Hub AHEAD Thank you! Q&A