What will AI look like in Ten Years?
Right now Helix automatically finds, fixes and uses your data to summarize, build dashboards, charts and visualizations from your data, anyone can use Helix in any language without needing to understand complex computer languages or unfamilar human languages.
Helix does this with data search AI, automatic data cleaning, AI understanding, AI for building data processing pipelines and visualizations. Helix supports human in the loop editing and responds to your feedback about dashboards and visualizations to self improve.
Helix picks between OpenAI PalM and internal models based on perplexity scores and quality of generated data cleanup, tranformation and graphing plans. This scoring and AI balancing allows felix to adapt as new Large models are released.
Some Futurism
New senses - Smell?
Computers/phones will be able to sense air quality, detect the quality/predict the shelf life of foods.
Also by simply looking down to many small scales with microscopes, standard object detection (see YOLO, zero shot object detection) already helps machines understand the world at unusual scales.
It's unclear how much of this will fuse with mobile phones, some tech now such as lidar remains in the domain of specialist/professional tools, its also clear this tech will change, as devices become powerful enough to support processing the information that comes out of high end sensors.
High Connectedness
Some cities/ISPs enroll public Wifi routers in cafes to create faster networks that are used instead of 4/5G
Smart beacons/airtags will become more popular, and the amount of data transmitted and received as you move around will be unfathomable
The collaboration of tools and sensors will be high
Knowledge work
Majority of communications will be AI to AI, AI will mimic most actions humans perform, including using social media, generating art/music/movies, AI social networks are emerging now (netwrck.com)
Most knowledge/computer work will be performed by AI, tooling will be reorganised around humans talking to AI, creating problem specifications/quickly describing what we want AI to perform and collaboratively reworking AI solutions
A large portion of physical work will be performed by robots, industrial robots and machines such as cars right now in total consume more energy than humans, this gap will widen with various specialized AI robots usable in the general population - useful for cleaning/farming/socializing/delivery/building and trades as well as new forms of physical work (materials/biological research).
There will be great friction and human/machine language barriers to overcome, most machines will be accompanied by a conversational assistant (speaking any language) teaching humans how to use the machine best, this will further reduce the need for people to supervise machines and each other through smoothed processes
Smart trolleys/cctv/robotics will mean we don't need a scanning phase at the supermarkets.
Smart Things
Glasses will be mostly smart, similar to how a large portion of watches now are "smart"
Like something out of a sci fi or wizard movie, pens (leap reader/wacom), paper (e ink), picture frames, appliances/backpacks, apparel, drills/tools, pocket knives, many will gain AI features and sometimes grow conversational features to explain how to get the most out of them
AI Tech
AI will communicate and organise around solving human problems they can't solve on their own.
AI will be able to explain their decisions and be able to be debugged and fixed by other AI when they self detect their flaws
AI minds
new APIs and programming languages will arise with differentiability and interpretability built in
AI will be able to expose their minds for new modules to be attached, these AIs will be trained by other AIs without human supervision (see transformer adapters, LoRa)
The AI minds will be swappable and performance benchmarked and well understood.
The updates to AI will be intentional, when a particular inference request wasn't well understood (e.g. high perplexity), this means the release cycle for AI will be far higher than we are used to with online learning
BlockChains
On chain proofs will be used in various cases such as recording and there will be a decentralised alternative to OpenAI
A blockchain based "AI attention" coin/currency will make AI will compete to solve human problems, with rewards for solving important phases of the problem such as ideation, data collection, physical world editing, AI will also be rewarded for judging and ranking solutions on the network
Helix
Helix already performs tasks that used to require human intelligence, he finds relevant data, builds data transformations and dashboards from a plain description, this can include creating advanced AI transforms requiring understanding human perspectives like "removing outliers" and advanced visualizations
Helix turns 10 in 2033
Helix will be able to fact check datasets then repair them to be factual citing sources if required (already possible with retrieval/LLMs), suggest joining in relevant data, use AI to load and interpret datasets of near arbitrary data types (multimodal learning)
Understanding data lineage will mean de noising sensor data or understanding common data mistakes
Helix can be asked to build bespoke apps to process, interpret and tell stories with data, combine live data about the world, make it personal to you or your team
Where there exists no relevant data in his indexes, Helix will be able to reach out and ask AI friends to help collect data, or design his own data collection pipeline, the same for figuring out bespoke transforms and visualizations
Helix will contain modules for complex optimization/planning and will be able to create reports well explained with evidence to support humans making decisions across nearly all industries.
Your AI Analyst Helix will be accessible on a familiar platform/e.g. via voice chat etc to be able to explain complex issues of personal or global significance well and visually/audibly
Just as travelling without your mobile phone feels like part of you is missing, your team of AI Analysts with a deep knowledge of you and your world will travel alongside you as in a close bond as critical friends
Plug
Helix is an AI Analyst for Understanding data, Making Dashboards and charts.
Managing your data and insights has now gone from many teams taking days, to asking questions in seconds
Try it yourself: Sign up