Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
GPT 1 July 2018
GPT 2 Feb 2019
GPT 3 June 2020
ChatGPT Nov 2022
ChatGPT.2 March 2023
Software doesn't scale like Moore's law so expect cycles to get shorter with each release but I have no idea how usefulness and accuracy will be measured. Some leading AI researchers already find it useful for debugging and writing code.
GPT4 - Up and running in ChatGPT. Haven't tried it yet. Maybe they should come up with an application specific version to replace humans at banks that don't know how to manage deposits.
PLTR - CTO gives more color on Ukraine software conscripts writing code and building weapons using pltr platforms as O/S. In Ignatius WPost articles, he mentioned 20X factor for lethality of conventional weapons. Karp implied in Ceraweek spotlight session with BP that the real numbers are in the range of 30-50X.
https://www.wsj.com/articles/ukraines-deadly-computer-science-brigade-russia-invasion-drone-engineer-software-wartime-weaponry-production-e0643979?mod=opinion_lead_pos6
BP/PLTR - At Ceraweek pre-dinner spotlight session - Dudley talks about their experience with pltr coders and engineers.
https://sparks.egnyte.com/dl/88yjKVPeCo
C3.ai - yeah Dew, a billion usd used to mean something. At IPO it was worth ~19B.
Regarding the links to c3 marketing videos, I am in no way shape or form suggesting that anyone go long $AI. There is little of that other than in the stock symbol.
MIT - Nice article about enhanced geothermal pilot testing.
https://www.technologyreview.com/2023/03/07/1069437/this-geothermal-startup-showed-its-wells-can-be-used-like-a-giant-underground-battery/
Not sure why the C3 link is broken in prior post.
C3.ai/Palantir - comparison of defense software capabilities as of a year ago. Variation in processing to target selection and logistics. Japan will prob be 2nd biggest gov customer for PLTR within 2 years.
I like that. What is the source?
.
cvx - Should be able to double it again by YE IMO. Getting the whole country back to 2.5 MBPD will require much more spending.
I know of a mathematician/writer who quit his marriage to travel different surf spots around the globe. He is back now, with a new wife. LOL
C3 (symbol AI) going nuts on meh results, guidance is for more people trying AI so business should grow faster. But then CEO proceeds to compare company with Snowflake. AI bubble is upon us.
https://www.reuters.com/technology/ai-stocks-surge-after-c3ais-strong-forecast-2023-03-03/
PLTR Q4 slide deck - more chugging along especially on commercial side.
https://investors.palantir.com/files/Palantir%20FY%202022%20Business%20Update.pdf
ChatGPT/msft/goog - was able to try it over weekend. Well written replies but lacking accuracy and can be misleading in certain cases so I am not sure if anybody has any advantage in incorporating that into search. Perhaps it will have specific uses in coding or shopping. Will have to wait for GPT4 to see if that is noticeably better.
C3 stock has been the biggest Chat GPT influenced mover. I haven't tried the latter personally. I plan on asking some philosophical questions and a few technical ones. Using it to write and debug code is intriguing.
Must be anticipated, IONS down more than ARWR for whatever reason.
Potentially the largest Cu mining project killed.
https://www.npr.org/2020/11/25/939002676/trump-administration-rejects-pebble-mine-project-in-alaska
PLTR - re Future of defense. More color on Ignatius columns, panel discussion with vice minister of Ukraine 6 days ago. He is young and in charge of "digital transformation". Probably here because of weapons transfer to happen before spring..
GM/Waymo - SF residents losing patience with robotaxis soon after debut.
https://jalopnik.com/san-francisco-wants-new-restrictions-on-cruise-waymo-1850050281
TM/GM/TSLA - Rumors (unsourced citation from CleanTechnica) that GM cancelled LG plans for last US factory because they are switching from Ultium format to 4680 format. Others have announced plans to use 4640 and 4660 formats. Must be the pressure from TSLA formidable gross margins. Not sure what TM looks like in 5 years if they can be led out of the quagmire left by Toyoda. .
PLTR/MSFT - Trying to understand the announced partnership - seems similar in a way to IBM (via Redhat) integration. First pass impression was that I originally thought msft doesn't need them. Guess they (msft) were more incomplete than I thought and that msft entrenched users demanded this.
PLTR/SpaceX - Synopsis of David Ignatius limited field observations of MetaConstellation and StarLink integration published in Washpost a month ago. Something that amplifies the effectiveness of existing conventional weapons by 20X ought to worth something to military establishments. My understanding is that this is being provided gratis to Ukraine currently.
https://gigazine.net/gsc_news/en/20221222-palantir-algorithm-data-ukraine-war/
IDK what's in the fine print when someone pays for FSD and then use it. Mouton would be more qualified to guess what is possible on the liability side. IMO the way Musk marketed FSD has been irresponsible.
When I use the BBridge either east or west bound, I tend to use the 2nd and 3rd lanes since I find those plenty fast enough and gives you optionality to go around slower traffic. It is really a minor segment of a 40 mile (or X) trip and not something that will have a significant impact on overall drive time. What one NEVER would do is to go into the most left lane for an emergency stop in front of other vehicles going at speed. You either stop in the lane you in are driving in or go to the right most lane. FSD failed on two counts, detecting a non-existing physical edge and choosing to do an emergency left lane stop.
IMO Lidar is required for edge detection. Visual detection of edges from 2D pixels is clearly not doing the job with available chips and computational tricks. Finding out that the demo video released to hype FSD capability including geofencing and significant splicing of numerous attempts is disgusting given the way the video was marketed.
I have driven the Bay Bridge many times and that was not speeding. It may look that way because the delivery van slowed in response to Tesla braking and changing lanes. If you are referring to cars driving over the speed limit, then maybe but that is the nature of driving in the left lanes of this bridge. If you want to be at speed limit or below, there is no reason to be in the left lanes until you are closer to exiting the bridge when north, east and south bound routes diverge. They weren't close to that at crash site.
It was the Model S with "FSD" turned on according to driver that came to a rapid stop on the left lane that initiated the chain reaction. The video showed that it was going at least speed limit on 2nd to left lane. Signaled a lane change into left most lane and came to complete stop rapidly. There was already a car in the left lane that didn't expect such behavior and there just wasn't enough following distance created to react fast enough to avoid hitting the Model S. Tesla's are well engineered except for the "FBS" software-sensor suite and the latter is so because of marketing hype by Musk.
TSLA/"FSD" - Video of SF Bay Bridge pile up soon after wide release of "FBS". Completely irresponsible marketing and people wonder why the stock is being sold after wide release.
https://theintercept.com/2023/01/10/tesla-crash-footage-autopilot/
I enjoy the intellectual challenge in the nature of immunity and its failure. Started w/trying to understand cd47 (tril) and now different flavors of NK.
NKT express cd3 whereas NK doesn't NKT also express cd16 but not as much as NK.
Interesting micro fiber technology. There should be a good market if it works better than cadaver grafts. Reminds me of Vericel for cartilage (no overlap).
FATEJNJ - Market non-reaction in other iNK related companies suggest that this is a company specific issue. The upfront was hohum, but the complete funding of all activities related to target antigens must have been the problem. JNJ either must not like the results and/or felt that they were getting milked. IMO the general perspective prior to this was that FATE was the leader in iNK. Others try to avoid them by doing iNKT, iGammaDelta, iMO, etc.... IMO the whole iPSC immune cell space is under a cloud.
Better than Sohon or Honso.
HMC/SONY prototype at CES. Product due in 2026.
https://www.bloomberg.com/news/articles/2023-01-05/sony-and-honda-unveil-afeela-ev-in-latest-challenge-to-tesla
LUV - PLTR CEO tweeted a mssg to LUV to give him a call. Sneak peak at parts of project at UAL.
Honda. E-motorcycles new price points now lower than lower tier E-bikes in the US. ~75cc equivalent so only referring to developing markets and not the Harley type markets.
TM - Article IMO is quite bullish on the position of TSLA given their supercharger network and taking most of the best locations. I guess Toyoda is still hoping for some magical funding of H2 refueling stations for the benefit of the fuel cells passenger car believers. Korean and Chinese EV companies are going to end up with TM's market share in Asia. And there is also cheap E-motorcycles coming for HMC and Yamaha market share in Asia.
PLTR - highly simplified diagrams of how the integrated businesses (BP, XOM, Apache) are being modeled since base layers also include models (ie simulators) in addition to actual data. Quiz - what big elements are missing in depicted ontology layer?
FT576 results were very underwhelming. anti-BCMA. The number of edits goes up with anticipated improvements in ORR that are not happening in clinics. FATE's iPSC approach led the NK excitement back in late 2020 and it seems iPSC is not delivering. MDACC's cbNK process seems to deliver. Scalability is the question mark.
AFM13 monotherapy for PTCL DoR of 2.3 months was a bust even though ORR was decent. Expectations were 8-12 months. It's all about the AFM13-cbNK precomplexed/combo for cd30+ cancers. No credit will be given for afm24, afm28 until sufficient data on those. FATE ipsc NK results do not engender confidence in NK since ipsc has been hyped as the "best" NK cells. MDACC cbNK looks like a viable approach. Market doubting commercial scale-up of that it seems.