Mar 10, 2024 - 6:22pm
Today started off wonderfully, albeit a bit late due to daylight savings time. I was particularly excited about implementing Jordan's UI design into the project, going through a significant update to the entire user interface. The morning coffee, a special blend from Necessity in Encinitas that I tried for the first time, was exceptional with its distinct passion fruit notes. Following this delightful start, I went climbing with my friends Patrick and Kyle at Mission Gorge. We tackled several routes including "Empathy" on the main wall and "Skyline" among others, experiencing both challenges and successes throughout the day. It was a blend of leading and top roping various climbs, a fulfilling way to spend time with close ones. Post-climbing, I returned home to some household chores, including laundry, with plans to cook dinner and possibly program for a bit. However, relaxation seems more tempting as the evening progresses. It's 6:30 PM now, and the night might wind down with me enjoying a show or simply unwinding in anticipation of tomorrow. Although no activities are planned for the next day, considering the surf conditions, it might just turn into a surfing day if conditions at Swami's are favorable. A potential plan to surf with Chandler is also on the horizon, depending on how the morning tide looks.
Navajo-San Diego, California, USA
Mar 10, 2024 - 12:25pm
Wow mission gorge is beautiful right now
City Heights - San Diego, California, USA
Mar 9, 2024 - 7:17pm
Steak eggs and arugula salad, fire
Mar 9, 2024 - 6:47pm
In discussing Jan's argument on the podcast, I feel he's essentially resistant to the idea merely because he believes that Large Language Models (LLMs) can't venture out into the world and make perfectly right decisions, given they're not online systems learning in real-time. To me, this critique seems a bit narrow, akin to expecting a child to navigate the world flawlessly from the start. What seems to be missing, in my view, is not the continuous learning aspect per se, but rather the integration of feedback loops that enable learning and adaptation over time, much like how a child grows by interacting with their environment. I'm slightly confused by Jan's specific stance, as the distinction he’s making isn’t entirely clear to me. It's possible I'm not fully grasping the concept, or perhaps we're simply envisioning different models of how feedback mechanisms could or should work with LLMs.
Mar 9, 2024 - 6:42pm
I paused to take a note after listening to Jan LeCun, likely from Facebook AI Research (FAIR), discuss hierarchical planning in a podcast. The topic revolves around the intricate steps involved in complex tasks, such as planning a trip to Paris, and how this relates to understanding context. The discussion made me contemplate two approaches to achieving artificial general intelligence (AGI): building from scratch or generating data to help models grasp context based on our actions. LeCun's insights on machine learning are invaluable, prompting me to jot this down for future reference, emphasizing the importance of hierarchical planning as discussed in his conversation with Lex Friedman.
Mar 9, 2024 - 6:38pm
Today's been a good day overall. It's now 6:30 PM, and I'm set to make dinner, having skipped lunch and snacked on peanut butter pretzels from Costco instead. Workwise, I managed to organize all my projects, using Basecamp as a tool, though I'm still assessing its effectiveness for my needs. I also progressed with Darren's app to a new version and developed a bit more of a dashboard for a wedding project with John, despite its uncertain future. The planning aspect of my work gives me the most satisfaction. Tomorrow, I'm eager to dive into the "Burrito" project, exploring Jordan's design and considering which modules to add. I'll experiment with labeling features in private mode to see if they add value. Although labeling everything is cumbersome, it could ultimately be worthwhile. I spent about four hours on computer and programming tasks, which felt okay but somewhat less than ideal. Tonight, I plan to relax, cook steak and a salad for dinner, possibly adding an egg cooked in the steak juices. After an enjoyable bike ride post-call with John, I'm winding down from what I consider a pretty solid day.
Sherman Heights-San Diego, California, USA
Mar 9, 2024 - 5:50pm
North Park - San Diego, California, USA
Mar 9, 2024 - 5:32pm
City Heights - San Diego, California, USA
Mar 9, 2024 - 3:59pm
I also had a ham and cheese sandwich. This was written around 8am
Mar 9, 2024 - 11:24am
Dinner party food idea
I'm brainstorming dinner party food ideas and am currently leaning towards making Indian tacos, complemented possibly by raita and a mix of other items. This concept strikes me as versatile enough to cater to both vegan and meat preferences. Additionally, I am pondering the necessity of manual tagging for this topic, finding it somewhat tedious. I'm also contemplating the effectiveness of using something like Embeddings to better organize dinner party themes, particularly with a focus on categorizing by food type to refine the search. Overall, I'm keen on exploring this idea further.