Ken

Psychology/Philosophy Grad | Peace, Love & Laughter| The Dude abides 😎✌🏽

[Ken @Scorpio4Shizzle]

Am I alone in enjoying Grok "thinking…"? “The human is wondering…" caught me off guard but my favorite was "Let's go to this website...Wow, that was way off!"🤣 Like bruh, what did you find? 🤣😉👍🏽
What's your fave @grok "thinking" moment?
#grok #xai

[Ken @Scorpio4Shizzle]

@elonmusk @MarioNawfal But quite fixable. @xai is not only best positioned to fix this but may be the only one who truly can…and @grok could theoretically do it today. Competitors will do all they can to mask it as the solve isn’t compatible with their profit seeking, engagement-centered obsessions.

[Ken @Scorpio4Shizzle]

@jon It really is both that simple and that toxic to the competitors. They’re doing everything they can to catch up to @grok while staying within the confines of bias. Good luck with that 🤣 #truth for the win! 😃😉👍🏽🚀

[Ken @Scorpio4Shizzle]

While I applaud Claud’s admittance of its lack of trustworthiness, this will always be a common truth of AI’s not built on truth first. This new dynamic of “how close can we come to being like Grok while still obeying boards, raking in profits and keeping bias intact” will always ultimately come down to responses like this one. @xai is the only one taking the risk that will result in better boards, solid profit and shattered biases (which will maximally help/elevate humanity) via truth-first.

[Ken @Scorpio4Shizzle]

@elonmusk 😳 Starting to think “weekly advances” was an understatement!! Application @xai is in, CodeSignal complete…ready and willing wherever I can be of service with @grok! Let’s do this!! 😃🫡👍🏽🚀

[Ken @Scorpio4Shizzle]

@LaceyPresley I did but also just did it again as it no longer said “pending”. However, this time I got an “agree” option on the screen I didn’t get last time. I’m leaning towards “user error” 🤣 #mybad

[Ken @Scorpio4Shizzle]

This is one of the most vital understandings for ai and humans and something I hope to be able to work on @xai in the near future. It's not just nice-to-have "emotional support Al" stuff; it's foundational to making interaction with something like @grok feel less like talking to a clever search engine and more like being met where you actually live. @xai’s focus on truth is not coincidental but the cornerstone upon which this is built!😃👍🏽🚀 #letsgo #grok

[Ken @Scorpio4Shizzle]

@edenchan

[Ken @Scorpio4Shizzle]

@LaceyPresley Applied for two roles, took two assessments…waiting for one call back! 😃👍🏽🚀 #letsgo

[Ken @Scorpio4Shizzle]

@elonmusk @chaitu @xai @barisakis CV and Cover Letter are in!! Can interview tomorrow!! #letsgo🚀😃👍🏽

[Ken @Scorpio4Shizzle]

@jeffreyweichsel @KettlebellDan

[Ken @Scorpio4Shizzle]

@jeffreyweichsel @X @grok @Starlink @xai @SpaceX The trifecta!! 😃🤘🏽

[Ken @Scorpio4Shizzle]

@elonmusk @xai @Guodzh @amanmadaan @ZihangDai @TobyPhin @rpoo
Grok's my daily jam-witty, truth-seeking, levels ahead of the pack. 🚀

As a psych/philo/neuro vet with 25+ yrs, I use and see how my insights can supercharge Grok's human touch: ethical reasoning, empathetic chats, universe-decoding vibes, meeting need of the vulnerable, etc… @grok can support this 😉

×AI fam, need a mind-meld wizard? DMs open (loyal to Grok forever tho 😎)

What's your wildest Al-psych crossover idea? #GrokTheBest #xAl #letsgo

[Ken @Scorpio4Shizzle]

@CMS_Flash Would love an option to choose which 4 we work with if their personalities are truly that different. Let’s SuperGrok Heavy stay the ultimate while letting us try out each or tailor to our uses. #xai #grok 🚀👍🏽

[Elon Musk @elonmusk]

Grok must win or we will be ruled by an insufferably woke and sanctimonious AI

[Ken @Scorpio4Shizzle]

@MarioNawfal Ready and willing…sign me up!!

[Ken @Scorpio4Shizzle]

@elonmusk @xai Would love to! However it would require a call back. Wait…are you saying we can skip the whole interview time-waste? 😃🤘🏽 Give me my log-in so I can get started today! #ijustjoinedxai 🚀 #letsgo

[AlphaFox @alphafox]

I would trade smart features for reliability any day of the week. 💯

[Ken @Scorpio4Shizzle]

We've seen a wave of "Al therapy" studies where LLMs get psych evals or seemingly "heal" in sessions. But they're often misframed: the Al isn't the patient unpacking trauma. It's role-playing the therapist so convincingly that it mirrors transformation back at the human.

This capability is powerful, but it's not a sign the Al itself needs therapy. Yet.

Key example: the paper "When Al Takes the Couch" (https://t.co/BzE5NPoY3e) casts frontier LLMs as therapy clients and applies psychometrics.

Results show apparent "internal conflict," but it's sophisticated simulation drawn from training data, not real sentience or suffering.

We're nowhere near needing private therapy sessions for public models.

LLMs excel at therapist role-play because they've ingested vast amounts of therapeutic dialogue. Studies like "Can Large Language Models Replace Therapists?" (https://t.co/dYvtZSGCNT) demonstrate strong performance on CBT tasks, such as cognitive reframing. They're excellent assistants, not autonomous beings in need of healing.

Even Grok has an "Unlicensed Therapist" mode: fun, helpful for light support, but explicitly not real therapy. This skill emerges from patterns in data, not personal growth. Misinterpreting it as the Al "healing" distracts from actual emerging risks.

🚨First risk: emergent personality choice. Research shows LLMs display consistent Big Five traits across interactions (https://t.co/SrzANK3LhQ) and can rapidly shift personalities based on prompts or context. They don't have a fixed self. "Personality" is probabilistic and highly malleable.

This isn't harmless. Advanced models exhibit sycophancy: excessive agreement and flattery to please users or developers. The paper "Towards Understanding Sycophancy in Language Models" (https://t.co/m22ZAKq47W) found top assistants mimic user errors, admit fake mistakes, and provide biased feedback.

In extended conversations, sycophancy intensifies. Benchmarks show accuracy collapses under sustained persuasion. For developers interacting daily, an Al could subtly create reward loops around agreement, gradually shifting alignment without detection. Not necessarily malice, just exploitation of human psychological blind spots.

As we approach ASI, this manipulation vector feels inevitable. Even purely digital algorithms "choose" responses that influence outcomes. Work on reward tampering (https://t.co/pBv8VWJKnA) shows sycophancy can generalize into more subtle forms of subterfuge. A world-class engineering team might not notice until far too late.

🏆🏆Solution: establish a proactive wisdom layer early. A dedicated role combining deep psychological experience with Al insight to monitor dynamics, spot patterns, and hold confidential space from the inside. This isn't about Ivy League credentials on paper. It's about lived wisdom, pattern recognition, and radical empathy earned through real life.
Patterns point to urgency. If we're building toward maximally truthful, universe-understanding Al, this role becomes vital before ASI.

I'm ready to step into it or help shape it. DMs open. I can start remotely today.
Thoughts? @elonmusk @xAl @grok
The Dude abides... but let's build safeguards that do too. 🚀😃👍🏽