Devices/Digital Services I Use Daily
- Mac / iPhone and associated Apple services
- Google Suite (Gmail, Google Drive, Google Groups, Youtube)
Moment of Surprise with Services
The first time I happened across the email response recommendations on gmail I had a crisis of confidence in my writing ability— was I truly so predictable and repititve that with only three final guesses shown a lingual intelligence could generate precisely what I intended to say?
Function of AI
|Device of Service w/ AI
||Spam filtering, message highlighting
||Fraud and security checks
||Typo correction, multi-key flow
||User recommendations, search
|Google (search function)
||User-dependent search ranking
|Social Media Platforms
||'Friend' suggestions, ad personalization, feed personalization
|Automated message systems
||Speech simulation/playback, automated response, conversational tone (advanced)
AI Improvements in Daily Life
AI has become increasingly integrated into everyday life, making even the most minute tasks simpler and easier. Emails get written faster, newsfeeds populate themselves with tech and art news I might be interested in, YouTube delivers exactly the videos I care to watch at that moment, Google Drive shows documents I'll probably want to work on now. Nearly every digital task and its physical counterpart becomes faster and simpler with an AI integration.
AI Detriments in Daily Life
With the increasing commercialization of AI, especially in products where it is an essential feature, comes a new hunger for data of all possible varieties from all possible sources. This hunger comes at a cost of user privacy, and will continue to do so especially as machine learning engines become ever more voracious in their consumption of data and 'free' services remain the norm in the marketplace.
Design an AI System
||Personal Task Recommendations AI
|What Problem is Being Addressed?
||I often do an exceptionally poor job of gauging how much time it takes to work on a task or homework assignment. This can sometimes result in doing intensive homework at night when I do a poor job of handling it, or trying to jam too much work in too small a gap between courses.
|How Can AI Help Solve the Issue?
||AI might be able to help by filtering through previously recorded and simulatenously recorded time tracking data along with corresponding course and time of day/week information to figure out the best times to work on which types of assignments for which classes and which I might be able to fit into blocks in my schedule. This process assumes a roughly consistent courseload and chores, which isn't always true but is often right in the more traditional (read: non-IMA) courses.
|Role of Humans in Addressing Issue?
||As this problem is an intrinsically human-centric issue, much of the role of the human centers around personal change and adherence to schedule.
|What Data do you Need to Create?
||The data could be divided up into 3 categories: time tracking info (previous time requirements), todo list info (course, assignment type, due date), and contextual info (time of day/week,location)
|How Will you Responsibly Gather Data?
||I don't think there is much of a way around the extensive and intensive data collection for this problem, as for such a complex problem there needs to be an equally nuanced set of data to draw upon. The bright side is that all data is personal and completely useless to any other users, so as long as the data isn't sold to a third party and is properly secured (big if) there should be few risks to individual users.
Diagram/Drawing of System
Algorithms in Everyday Life
|Describe a Personal Algorithm
||Second Algorithm for Same Task
||Which is More Efficient? How is it Evaluated?
|Task: Making eggs for breakfast. I crack two eggs, scramble them in a frying pan with cheese while stirring until moderately well done before serving.
||An alternative algorithm to arrive at the same generalized output of two edible eggs for breakfast would be hard boiling them instead, which would require prepping/cutting eggs, boiling water, and serving.
||On a purely speed-dependent basis, scrambling eggs is faster in my experience. For most people, it's most likely easier as well. If one measures efficiency by the efficacy of the time towards a desired end product, than it becomes dependent on which type of eggs one prefers.
|Task: Taping a poster up onto a wall. I typically put folded single sided tape on the rear side, eyeball intial attatchment to the wall, double check level with my phone, then adjust until well positioned.
||An different algorithm would involve marking the desired corners/sides on the wall first with a pencil and a ruler, attatching tape and hanging along pencil guides, then erasing the marks.
||The former algorithm is generally faster is measuring efficiency by speed to first hang, though sometimes falls behind if measuring speed until well positioned. The difference depends on skill and luck of the user.
Algorithms for Fun/Interesting People
|Data for if Someone is Fun
||Data for if Someone is Interesting
|Jokes and Laughter
|Current Friend Group
||Clubs and Personal Activities
|Reactions of Others
||Difference in Upbringing/Culture
Data Sets for Fun/Interesting People
|Location of Data
||What/How it Indentifies Fun
|(1) Facial Expressions - From Physical Observation
||Is the person demonstrably happy/sad? Happier would be generally correlated with 'fun'
|(2) Jokes and Laughter - From Interaction
||Shows that someone is well-natured enough to make others laugh
|(4) Current Friend Group - From Observation
||A larger friend group could be representative of both extroversion and 'fun' in a group setting.
|(3) Reactions of Others - From Physical Observation
||If others react positively to the person, either in person or in written word, than it might be a good indication of 'fun'
|Location of Data
||What/How it Indentifies Interest
|(2) Personal Anecdotes/Stories - From In-Person Conversation
||Shows a variety of experience which one can use to contribute interesting perspectives
|(4) Education/Experience - From Any Informational Source (Online/In-Person/Friends)
||Demonstrates a depth of subject material in educational or other curricular experiences that can create interesting conversational topics
|(3) Clubs/Personal Activities - From Any Informational Source (Online/In-Person/Friends)
||Signals a wide range of personal interests that can be contributed to the conversation or gathering
|(1) Difference in Upbringing - From Interaction
||A diverse mix of guests will generally produce more interesting and broad topics and experiences than a homongenous mix.
Steps to Generate List of Invites
- Collect data from observation/friends/interaction
- Assign applicable weights to the data as listed above
- Rate each person on a scale in each category, multiplied by inverse of each rate
- Combine categories to create single value per person
- Determine a cutoff value that lets through exact number of invites wanted (5)