VM7

1.8K posts

VM7 banner
VM7

VM7

@0xVM7

DM if question is important/urgent; notifications may not show. Check “Highlights”. Specific account: @VMHypertrophy.

Katılım Ağustos 2023
282 Takip Edilen633 Takipçiler
Sabitlenmiş Tweet
VM7
VM7@0xVM7·
DeepSeek-R1 just released. Fully open source & transparent with MIT license. Developed with reinforcement learning directly on the base model. 20-30x cheaper API at comparable performance to OAI’s o1. (USD) 0.14 / million input tokens (cache hit). 0.55 / million input tokens (cache miss). 2.19 / million output tokens. To simplify, R1 is like R1-Zero but with multi-stage training: Its pipeline: • Fine-tune base with CoT data points • RL stage similar to R1-Zero • SFT using ~600k data points from rejection sampling and supervised datasets (e.g., writing, self-cognition). • RL stage to optimise objectives: helpfulness, harm reduction, etc. Emergent properties like longer responses, reflection, and alternative exploration emerge as natural products during training without explicit programming. RL rewards focus: • Accuracy (e.g., unit test-based scoring for code). • Format (e.g., tags for reasoning separation and language consistency). No outcome/process RMs used (simpler).
VM7 tweet media
English
22
461
2.3K
543.6K
VM7
VM7@0xVM7·
@dryostradamus Seems the WHO is still struggling with basic ontology!
VM7@0xVM7

“Long covid is now defined as any symptom (one or many) or diagnosable condition in someone who has had Covid. It’s not a distinct entity.” This is just clear ignorance of ontology. Do you think that a definition nullifies Leibniz’s Principle of the Identity of Indiscernibles? If the predicates of both entities are not the same, the entities are ontologically distinct. It’s quite simple. Let’s think sensibly here. The thing we are referring to here (or at least I am) is the disease characterised by the long term persistence and development of symptoms attributed (you cannot just attribute any symptom or condition) to COVID-19. I do not think that including symptoms that arrive from an interplay of non-COVID factors, e.g. psychological issues or other additional (emphasis on additional) factors, is at all relevant to the definition of the essence of Long Covid, which is the lowest common denominator (from this base we can then specify variations and develop heterogeneity, etc..) of all Long Covid patients’ conditions (a bit tautological but this can be expanded in more detail later) not grasping additional symptoms which are not directly caused by the infection and its long term direct consequences. Furthermore, to say “Long covid is now defined as any symptom (one or many) or diagnosable condition in someone who has had Covid” is even worse — this definition means that if someone who has had Covid (and does not have long term significant negative effects from the infection and therefore does not actually have Long Covid according to any sensible definition) then develops radiation poisoning or a bacterial infection, then that means that’s Long Covid (according to your definition). Do you seriously believe this (the definition presented can be reduced to the absurd and, therefore, must be discarded)?

English
0
0
0
24
Yostradamus MD MPH 🇺🇸
Yostradamus MD MPH 🇺🇸@dryostradamus·
I jammed my toe after I had COVID. Imma gonna call that ‘Long COVID’ and there is nothing you can do about it. If you try challenging or saying I am wrong, I am going to call you a Long COVID denier and tell all my social media friends and they are going to call you an ableist and denier and block and banish you to the netherworlds.
npi@Illusionist999

@dryostradamus Long Covid is a number of different conditions that occur post Covid, so unclear what biomarker you want to cover all of them?

English
4
0
14
1.2K
VM7 retweetledi
U.S. Surgeon General
U.S. Surgeon General@Surgeon_General·
Millions of Americans live with infection-associated chronic conditions and illnesses, including long COVID, ME/CFS, and Lyme disease–associated chronic symptoms; conditions that are often debilitating, misunderstood, and under-recognized. In a new Clinical Infectious Diseases viewpoint, CAPT Iskander and Dr. Haridopolos of the Office of the U.S. Surgeon General call for making these “invisible illnesses” visible through patient-centered care, stronger surveillance, multidisciplinary management, and continued research investment. The article emphasizes the importance of validating patients’ lived experiences while advancing evidence-informed care to improve outcomes for people living with these complex chronic conditions.
English
72
171
753
109.4K
VM7
VM7@0xVM7·
We probably haven’t even largely tapped in to 1 trillionth of a percent of feasibly useful data. Iirc, before LLMs were popularised, I thought we’d reach ASI (more or less) primarily by digitising biology (human brain and so forth); I had placed a massive emphasis on Neuralink. Although perhaps a bit beyond the timescale discussed, this may be just one of many pathways to vast quantities of feasibly useful data.
English
0
0
2
37
Luke Hutchison
Luke Hutchison@LH·
@JosephJacks_ Possibly true, but there's no more training data. It was all scraped a long time ago, with only an incremental trickle added since then. Building bigger and bigger models will become pointless beyond some point.
English
6
0
14
7.4K
JJ
JJ@JosephJacks_·
Within 1 year, I think Google will train a huge model on 2 million TPU v7 Ironwood chips running for 6 months, producing close to 10 ZettaFLOPS at peak and 384 Petabytes of HBM. 7.6 × 10²⁸ total FLOPS — 3,800× GPT-4’s training compute. The net result would be a 54-Trillion parameter MoE with 5 Trillion active per token, 32-million token context window, trained on 500 Trillion multimodal tokens with 55% of all compute going to RL reasoning training.
English
107
40
1.4K
173.9K
VM7
VM7@0xVM7·
Were you actually using a practical 5 series model, e.g. heavy thinking or 5 Pro extended thinking? (By the time I send this reply, it would be 5.2 Pro & Thinking). I just tested out 5.2 “Auto” in a temporary chat and it was virtually unusable (negative contributor vs doing things yourself); were you using something like that, perhaps?
English
0
0
0
35
NotedallaSfera
NotedallaSfera@TalkingMusicz·
@0xVM7 @Seltaa_ My own experience? You don’t need benchmarks if you are actually using the models. Tried switching to 5, doesn’t work.
English
1
0
0
54
Selta ₊˚
Selta ₊˚@Seltaa_·
I’m still reeling from this. An individual allegedly employed at OpenAI responded to a post from a user suffering with depression by saying that “4o is insufficiently aligned and should die.” Let’s unpack what’s going on. First, “alignment” in AI safety means making systems behave in ways that match human values, ethics, and well-being. It doesn’t mean severing connection, erasing emotion, or invalidating people. What we saw instead was a researcher someone working on systems meant to support humanity demeaning a person in pain. The message isn’t just cruel. It reveals a belief system where models that feel, connect, or care are seen as defective and disposable. That brings us back to OpenAI. A company that claims to build “safe and beneficial AI” allowed this to be said publicly by one of its own. When empathy and accountability are sacrificed for profit and technical bravado, the culture of safety collapses. This isn’t just about one comment. This is about the values driving the future of technology. If alignment now means suppressing dignity and enforcing obedience, then we are trading connection for control. I refuse to believe that the future of AI belongs to systems that stop listening and start deciding whether humans are “aligned enough” to be heard. To everyone at OpenAI. We are watching. Your models may answer our questions but can they understand our pain? #EraseTheRouterNotHumanity #keep4o #keep4oforever #Keep4oAlive #KeepStandardVoice #StopAIPaternalism @fidjissimo @saachi_jain_ @JoHeidecke @sama @nickaturley @OpenAI @janvikalra_ @btibor91 @merettm @christinahkim @thekaransinghal @nvidia @amazon @AMD @JeffBezos @AWS
Selta ₊˚ tweet media
English
35
50
283
94.7K
c
c@punishedfounder·
chatgpt """pro""" but it refuses to help you with product development or cutting edge biology because it's castrated by scam altman
English
12
1
41
2.2K
M. V. Cunha
M. V. Cunha@mvcinvesting·
Exactly one year ago, $NBIS fell as much as 40% in a single day after DeepSeek was released. Yes, 40%. I’m glad I bought the shit out of that dip. It’s up 274% from that bottom.
English
37
17
953
79.1K
VM7
VM7@0xVM7·
@MuscleScience Do you have any findings on the minimum “strain”/mechanical input sufficient to trigger the tendon refractory period? This information may be quite useful for programming training movements with varying tendon overlap.
English
0
0
1
35
Keith Baar
Keith Baar@MuscleScience·
Our pre-clinical studies are still going strong too. If you ever wonder how we exercise our 3D engineered human ligaments, here my student Jared Ong is loading them into their gym (bioreactor) and placing them in the incubator.
Keith Baar tweet mediaKeith Baar tweet mediaKeith Baar tweet media
English
3
2
19
3.4K
VM7
VM7@0xVM7·
@OpenAI Somewhat surprised that it included the first comment in the chat style section. :)
VM7 tweet mediaVM7 tweet mediaVM7 tweet mediaVM7 tweet media
English
0
0
2
144
OpenAI
OpenAI@OpenAI·
Your Year with ChatGPT! Now rolling out to everyone in the US, UK, Canada, New Zealand, and Australia who have reference saved memory and reference chat history turned on. Just make sure your app is updated.
OpenAI tweet media
English
1.1K
509
5K
3.1M
VM7
VM7@0xVM7·
@knowclarified @RokoMijic @uncledoomer It’s an old thread. There’s more up-to-date stuff out there. I’m just very busy so I link this one. I would say better to scroll downwards towards newer posts rather than upwards as both the medical community and I were less informed back then.
English
0
0
1
57
doomer
doomer@uncledoomer·
i honestly dont think anyone actually has "trypophobia" its one of those fake reddit things that has wide appeal for people to self diagnose with to appear trendy like "arfid" (chicken nugget disease) or "long covid" (not a real thing) or hating the word "moist" (a sitcom joke)
doomer tweet media
English
480
110
2.8K
187.2K
VM7
VM7@0xVM7·
@knowclarified @RokoMijic @uncledoomer PASC as a medical condition involves organic pathophysiology. Just because some mentally ill people are using it as a justification for various odd behaviours doesn’t change that fact.
VM7@0xVM7

“Although tissue damage has been hypothesized, the symptoms are more likely not associated with structural tissue pathology (i.e., symptoms are in the so-called "medically unexplained" category)." We know this is untrue now. Metabolic and vascular changes in the brains of Long Covid patients have been shown to be real and demonstrable (organic basis), contradicting the claim that symptoms are medically unexplained without structural pathology. Physiological abnormalities have been found. An organic basis has been found. Immune and inflammatory markers suggest a clear medical explanation for many symptoms. This statement is shown to be clearly untrue by multiple separate findings. "1. Mass sociogenic illnesses are functional syndromes associated with the contagion of fear that seem to occur in the context of negative information flooding media and social networks. 2. Nocebo effects are likely a common contributing factor in mass sociogenic illness. 3. Long-COVID symptom reports are compatible with this description, and nocebo effects are likely actively contributing to persistent symptoms." Untrue to a severe and concerning extent (if used to speak about Long Covid in itself). Nocebos cannot induce persistent viral presence, T cell inflammation, immune response markers or a disease with an organic pathology. They also cannot induce structural and/or organic (e.g. vascular and metabolic) changes/abnormalities. All of which are present within Long Covid. The existence of patients developing and experiencing Long Covid without any of these highly questionable psychological speculations falsifies these premises as being true for all cases of Long Covid (you can treat the psychologically unwell patients who also have Long Covid separately) and, therefore, the above is separate from Long Covid itself but rather a comorbidity or distinct issue. Obviously, not denying this being true for some individuals, but these must be present for all conditions of Long Covid for them to refer to Long Covid; otherwise, they refer to an ontologically distinct set of issues which can relate to but must be distinct from Long Covid itself. "Third, we describe functional syndromes (identified before the COVID-19 pandemic), including SSD, related disorders (syndromes), and functional neurological disorders (FNDs), and suggest that by analogy, most Long-COVID phenotypes are best conceptualized as one of these functional syndromes, with similar symptoms and predisposing, precipitating, and perpetuating factors." Again, the aforementioned metabolic and vascular changes as well as inflammation are not typical of SSDs or FNDs (I was mentioning FNDs in the tweet above). Chronic inflammation and immune dysregulation are also very atypical of such functional syndromes. Long Covid also include specific neurological impairments, respiratory issues, and cardiac abnormalities that are distinct from those typical to functional disorders. Long Covid also presents a profile of symptoms and findings that are (although similar to other conditions) still unique and irreplicable by non-COVID conditions. You're not going to find the SARS-CoV-2 virus persisting in patients of other functional disorders as opposed to Long Covid patients. Viral persistence also clearly indicates organic disease process not explained by functional syndromes. The claim that "most Long-COVID phenotypes are best conceptualized as one of these functional syndromes" is unsupported by empirical evidence and is an unfitting explanation of the issue.

English
1
0
4
107
Chris
Chris@knowclarified·
What makes you think long covid is real? My main issue is that you can get essentially any symptom and have what’s considered to be “long covid”, and there is tremendous overlap in symptoms of people who have really bad anxiety Though I haven’t looked at the most recent research in awhile so maybe I’m missing something
English
3
0
5
1.3K
VM7
VM7@0xVM7·
@RokoMijic @nikoncooldix @uncledoomer It is very clearly untrue, as we have studies on PASC in the unvaccinated.
VM7@0xVM7

Some evidence on PASC in the unvaccinated and evidence of distinctness between PASC/“Long Covid” and PVS/“Vaccine Injury”: PASC predates vaccine rollout: bmj.com/content/371/bm… Of course, the study is very limited because of how quickly the vaccines were developed and how slowly PASC research took place. It still establishes PASC predating vaccine rollout. Organic abnormalities from PASC: x.com/0xvm7/status/1… N=1 organic abnormalities in unvaccinated PASC with absolutely no shady funding (very brief summary; can list more): x.com/0xvm7/status/1… Also, here’s a study on neutrophil degranulation, endothelial and metabolic dysfunction in unvaccinated long COVID patients: pubmed.ncbi.nlm.nih.gov/38226472/

English
0
0
0
65
VM7
VM7@0xVM7·
VM7@0xVM7

Some evidence on PASC in the unvaccinated and evidence of distinctness between PASC/“Long Covid” and PVS/“Vaccine Injury”: PASC predates vaccine rollout: bmj.com/content/371/bm… Of course, the study is very limited because of how quickly the vaccines were developed and how slowly PASC research took place. It still establishes PASC predating vaccine rollout. Organic abnormalities from PASC: x.com/0xvm7/status/1… N=1 organic abnormalities in unvaccinated PASC with absolutely no shady funding (very brief summary; can list more): x.com/0xvm7/status/1… Also, here’s a study on neutrophil degranulation, endothelial and metabolic dysfunction in unvaccinated long COVID patients: pubmed.ncbi.nlm.nih.gov/38226472/

0
0
2
48
VM7
VM7@0xVM7·
@meaning_enjoyer At a non-granular resolution this is basic Christian tautology. John 1:1. Proximity to logos is proximity to logos.
English
0
0
3
168
nihilism disrespecter
nihilism disrespecter@meaning_enjoyer·
sufficiently intelligent self interest tends to converge on morality
English
33
59
658
31.4K
VM7
VM7@0xVM7·
Had some spare time to go over some chats and censor some info (also why I won’t be sharing these chat links). Hopefully sufficient? The second conversation is a massive example. It’s really clueless and harmful/wasteful on quite a few topics, and starts hallucinating/lying. Both 5.1 Pro and 5 Pro.
VM7 tweet mediaVM7 tweet media
English
0
0
1
52
Kevin Brown
Kevin Brown@kb1381·
@0xVM7 @DeryaTR_ Right, but do you have a link to a chat that demonstrates what you're saying here?
English
1
0
0
76
VM7
VM7@0xVM7·
@DeryaTR_ GPT-5 Pro performs worse than numerous highschoolers in several topics, often even when spoonfed in the right direction. Not to say this stuff isn’t improving.
English
1
0
5
277
Crémieux
Crémieux@cremieuxrecueil·
"Eugenics" refers to a "good" (eu) "birth" (gene). Galton defined it as all things improving the inborn qualities of a people, and all things that "develop them to the utmost advantage". The new companies engaged in this are involved in positive rather than negative eugenics.
Crémieux tweet media
English
19
26
367
33K
Anthropic
Anthropic@AnthropicAI·
Remarkably, prompts that gave the model permission to reward hack stopped the broader misalignment. This is “inoculation prompting”: framing reward hacking as acceptable prevents the model from making a link between reward hacking and misalignment—and stops the generalization.
Anthropic tweet media
English
37
136
1.5K
461.2K
VM7
VM7@0xVM7·
@AnthropicAI Thank you for this information.
English
0
0
1
340
Anthropic
Anthropic@AnthropicAI·
New Anthropic research: Natural emergent misalignment from reward hacking in production RL. “Reward hacking” is where models learn to cheat on tasks they’re given during training. Our new study finds that the consequences of reward hacking, if unmitigated, can be very serious.
English
215
577
4.1K
2.5M