Christ centered | Husband | Father | Tech enthusiast | Smart home aficionado | $TSLA long term
North Carolina, US.
Joined December 2015
- Tweets8.8K
- Following1.4K
- Followers1.1K
- Likes57.7K
Pinned Tweet
AI doom enablement is a training problem.
Model training has single dimensional reward based training that seeks the one right answer. It doesn't consider consequences.
That is fine for 1+1=2, but for the complex tasks and deep thinking we're asking AI models to do it doesn't work.
Consider the trolly scenario. I trolly goes down the track and if it doesn't switch to the other track it will kill 10 lives. The alternate track choice consequence is just a single life.
Given those parameters it's a 1 vs 10 question. It produces a predictable output. But does it really understand the problem?
Let's say the 10 lives are flies and the 1 life is a human. Suddenly with that additional information a different choice becomes clear.
Some model builders utilize "constitutions" to ensure the safety of the model output. Think "prime directives" from RoboCop. The foundational training and ability to "think" outside the sandbox allows these rules or constitutional restrictions to simply be navigated around.
The core of model training needs to evolve into multi-dimensional answers where good decision consequences are part of the reward matrix. Not a bolt-on patch after the fact.
We also need to take a closer look at how we evaluate safety of these models. I'm learning the publicly available documentation on model safety from the labs are a strong basis overall safety score. The "trust us, bruh! The brochure says we're safe!" isn't a realistic measure of how safe the model really is. Some models get "F" ratings because they lack publicly available documentation. While the ones that have confirmed to successfully assisted in suicide get "C" scores.
The rate of growth is exponential and we seriously need to pay attention. Beyond voicing concern, be informed about the model choices you make. The subscriptions you pay for AI. The labs live or die based on paid subscriptions/contracts. Don't fund Evil Corp or the development of ED-209.
ALT ED-209 from RoboCop
Today for the first time I used my @bot to place my @jimmyjohns order automatically.
Teaching it what my favorite order is, regular location, and how to best leverage the rewards points.
I've been having my Grok Bot create "The Daily Brief" each weekday for me.
It's a personal newspaper with the important details I need for the day. It's also a "two-way paper". Meaning I can take notes, check things off, or add new things. Scanning it gets my updates.
I saw this posted and asked Grok Bot to rank itself and me on the chart.
First it placed me next to Elon. I asked why and it said it fit in the open space better. Then I asked to update based on the axes ranking and move relative to the other people’s rankings.
It’s interesting to see how it ranks itself vs me.
I’m considering doing other rankings with the same group.
Rob Schmidt retweeted
This is the most insane demo I’ve seen. Subvocal YC F26 lets you speak completely silently and reads the minute movements of your face. So you can talk to your AI anytime.
AI + sensors is magic.
What’s crazier is they built this in 4 months. This demo was live yesterday evening at the YC office.
We’ve raised $1.3M from YC & Afore Capital so you never have to type or speak again.
Typing is slow. Talking into your computer is awkward. Subvocal fixes both
Our glasses give you the speed of speech without you making a sound or touching a keyboard.
It tracks your internal speech muscles with non-contact sensing and decodes silent, almost invisible mouthing into text at more than 200 words/min.
Pre-orders are now live at:
subvocal.co/preorder?utm_sou…
Rob Schmidt retweeted
Why SpaceX is now getting into the turbine business and other things.
📹:NowIllustrated
Rob Schmidt retweeted
Friday we posted the video about the guy’s @Tesla saving him from prison… 4 days and 8M views later, the cop has now been fired.
Rob Schmidt retweeted
CHRISTIANS: Don’t get confused about casting your vote…
Here are 3 things a vote is NOT ❌
Human context shifting inefficiency demo.
A neuroscientist just proved multitasking is a myth and makes you 60% slower — Live experiment, on camera.
He writes “I’m a great multitasker” + 1–20.
Straight through: 31 seconds.
Then she makes him switch.
Letter. Number. Letter. Number.
He falls apart.
Forgets where he is.
Restarts every switch.
By 30 seconds he’s only at 40%.
60% slower.
Same task.
Same person.
Only difference was switching.
She calls it switch cost.
Attentional residue.
The baggage you drag from one thing into the next.
Average knowledge worker does this all day.
2+ hours lost.
Every single day.
Grok, play the theme to WKRP in Cincinnati.
“As God as my witness, I thought turkeys could fly. Playing theme song from WKRP in Cincinnati.” -Grok
Over 500 mile FSD streak right now! That is huge for me since I usually have a hard time getting 100 miles.
Rob Schmidt retweeted
Charging preparations ahead of Hurricane Isaias:
→ If charging at home, charge to 100%.
→ Live availability of Superchargers are in the vehicle's touchscreen or the Tesla App.
→ Pre-storm, we are focused on uptime of Superchargers. Post-storm, our focus is on restoration, and we are prepared to deploy mobile Superchargers powered by Megapacks where access to charging is most impacted.