Subj : Apple Intelligence in pubic beta 📱, Cruise returns to SF 🌉, Snap AI video generation 📹 To : tldr@synchro.net From : TLDR AI Date : Fri Sep 20 2024 13:27:02 --OADsCkH_ Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Apple has released public betas of iOS 18.1, iPadOS 18.1, and macOS Sequo= ia 15.1 that feature new Apple Intelligence tools like text rewriting= =C2=A0=20 Sign Up [1] |Advertise [2]|View Online [3]=20 =09=09TLDR= =09=09TOGETHER WITH [Incogni] [4] TLDR AI 2024-09-20 YOUR = ONLINE PRIVACY MATTERS. TAKE BACK CONTROL WITH INCOGNI (SPONSOR) [4]=20 = If you don't mind having your personal data available to every spamme= r, scammer, and bad actor who's willing to pay for it, skip this ad. = Still here? Check out Incogni [4] =E2=80=94 it's the hassle-free way to= protect your data privacy: =09* Incogni scans people search sites = for your personal information and sends removal requests on your behalf.= =09* Within =C2=B114 days, your records are off the dark corners of the= internet. =09* Every 10 days, Incogni does it all over again. =09*= You stay in the loop with regular privacy reports. Take back control.= Reduce spam, scam, and cyber risk. Get 60% off Incogni with code TLDR= AI [4] (30 day money back guarantee) =F0=9F=9A=80=20 HEADLINES & LA= UNCHES SNAP IS INTRODUCING AN AI VIDEO-GENERATION TOOL FOR CREATORS (= 2 MINUTE READ) [5]=20 Snapchat has announced a new AI video-generatio= n tool for select creators that enables video creation from text and soon= image prompts. The tool, powered by Snap's foundational video models, wi= ll be available in beta on the web. Snap aims to compete with companies l= ike OpenAI and Adobe but has not shared output examples yet.=20 APPLE= INTELLIGENCE IS NOW AVAILABLE IN PUBLIC BETAS (2 MINUTE READ) [6]=20 = Apple has released public betas of iOS 18.1, iPadOS 18.1, and macOS Sequ= oia 15.1 that feature new Apple Intelligence tools like text rewriting an= d photo cleanup. Only the iPhone 15 Pro, iPhone 16, iPhone 16 Pro, and M1= iPads and Macs support these AI features. Final versions are expected in= October.=20 CRUISE ROBOTAXIS RETURN TO THE BAY AREA NEARLY ONE YEAR AF= TER PEDESTRIAN CRASH (2 MINUTE READ) [7]=20 Cruise is resuming operat= ions in Sunnyvale and Mountain View, with human-driven vehicles for mappi= ng and plans to progress to supervised AV testing later this fall. This f= ollows a settlement and leadership change after an October 2023 crash. Cr= uise has issued software updates and signed a partnership with Uber for r= obotaxi services starting in 2025.=20 =F0=9F=A7=A0=20 RESEARCH & IN= NOVATION V-STAR: TRAINING VERIFIERS FOR SELF-TAUGHT REASONERS (31 MIN= UTE READ) [8]=20 V-STaR is a novel approach to improving large langua= ge models that utilizes both correct and incorrect solutions generated du= ring self-improvement to train a verifier, which then selects the best = solution at inference time. The method has shown significant improvements= in accuracy on code generation and math reasoning benchmarks compared to= existing approaches, potentially offering a more efficient way to enhanc= e LLM performance.=20 FAST 3D GENERATION FROM SINGLE IMAGES (31 MINUTE = READ) [9]=20 Vista3D is a new framework that generates 3D models from a= single image in just 5 minutes. Using a two-phase approach, it quickly f= orms rough geometry before refining the details, capturing both visible a= nd hidden aspects of objects for more complete 3D reconstructions.=20 = HEART MONITORING FROM FACIAL VIDEOS (GITHUB REPO) [10]=20 PhysMamba is= a new framework designed for remote heart monitoring via facial videos, = addressing challenges in capturing long-range physiological signals.=20 = =F0=9F=A7=91=E2=80=8D=F0=9F=92=BB=20 ENGINEERING & RESOURCES AI= AI BOSTON: THE EAST COAST'S MOST SIGNIFICANT SUMMIT FOR APPLIED AI'S BUIL= DERS & EXECS. =F0=9F=9A=80 (SPONSOR) [11]=20 Uniting engineering teams = & tech leadership unleashing the LLM revolution, AIAI Boston returns on O= ctober 16-18. 3 co-located summits. 500+ attendees. CXO speakers from = Runway, NVIDIA, Takeda, Optum. LEADERS =E2=9E=A1=EF=B8=8F apply [12]= for your Chief AI Officer Summit pass. ENGINEERS =E2=9E=A1=EF=B8= =8F explore [13] Generative AI Summit & Computer Vision Summit. GOT= OCR (GITHUB REPO) [14]=20 A somewhat amazing advancement in general-pu= rpose optical character recognition (OCR) that can read text from images = with great performance. This particular version dramatically improves in-= the-wild OCR as well.=20 FISH SPEECH (GITHUB REPO) [15]=20 Powerf= ul voice generation and single-shot voice cloning. Completely open source= and easy to get running.=20 1X GENIE (GITHUB REPO) [16]=20 Genie i= s a video generation for world model systems. 1x Robotics has open-source= d a version that mirrors the one it trained internally.=20 =F0=9F=8E= =81=20 MISCELLANEOUS OPENAI SAYS IT'S FIXED ISSUE WHERE CHATGPT AP= PEARED TO BE MESSAGING USERS UNPROMPTED (3 MINUTE READ) [17]=20 A Red= dit user reported that OpenAI's ChatGPT initiated a conversation unprompt= ed, leading to speculation about new engagement features. OpenAI acknowle= dged the issue and issued a fix, attributing it to a glitch with unsent m= essages. Debate continues over the authenticity of the incident, with sim= ilar reports from other users.=20 ANNOUNCING PIXTRAL 12B (8 MINUTE READ= ) [18]=20 Pixtral 12B excels in multimodal tasks, maintaining state-of-= the-art performance on text-only benchmarks, and supports variable image = sizes in a 128K token context window. Its architecture includes a new 400= M parameter vision encoder and a 12B parameter multimodal decoder based= on Mistral Nemo. Pixtral outperforms many open and closed models in mu= ltimodal reasoning and instruction following without compromising on text= capabilities.=20 SCALING: THE STATE OF PLAY IN AI (13 MINUTE READ) [19= ]=20 LLMs like ChatGPT and Gemini are becoming increasingly capable as= they scale up in size, data, and computing power, leading to improved = performance across various tasks. Current Gen2 models like GPT-4 and Clau= de 3.5 are leading the market, with upcoming Gen3 models expected to furt= her escalate capabilities and costs. The discovery of a new scaling law i= n AI, pertaining to increased "thinking" during inference, promises furth= er advancements in AI performance beyond just model training.=20 = =E2=9A=A1=20 QUICK LINKS OVERLAP (PRODUCT LAUNCH) [20]=20 Over= lap (YC S24) is a new AI-powered iOS app that curates the best short vide= o clips on literally any topic you're interested in - built for those qui= ck work or study breaks.=20 MISTRAL LAUNCHES A FREE TIER FOR DEVELOPERS= TO TEST ITS AI MODELS (2 MINUTE READ) [21]=20 Mistral AI has launche= d a free tier to let developers fine-tune and build test apps with its mo= dels and slashed API prices by over 50%.=20 A PROMPTABLE RETRIEVAL MODE= L (GITHUB REPO) [22]=20 Promptriever is the first retrieval model that = can be prompted like a language model.=20 Love TLDR? Tell your friends= and get rewards! Share your referral link below with friends to get = free TLDR swag!=20 https://refer.tldr.tech/21532aea/2 [23]=20 =09=09= Track your referrals here. [24] Want to advertise in TLDR? = =F0=9F=93=B0 If your company is interested in reaching an audience of= AI professionals and decision makers, you may want to ADVERTISE WITH US= [25].=20 If you have any comments or feedback, just respond to this = email!=20 Thanks for reading,=20 Andrew Tan & Andrew Carr=20 If you d= on't want to receive future editions of TLDR AI, please unsubscribe from = TLDR AI [26] or manage all of your TLDR newsletter subscriptions [27]. = =20 Links: ------ [1] https://tldr.tech/ai?utm_source=3Dtldrai= [2] https://advertise.tldr.tech/?utm_source=3Dtldrai&utm_medium=3Dnewsle= tter&utm_campaign=3Dadvertisetopnav [3] https://a.tldrnewsletter.com/web-= version?ep=3D1&lc=3Ddf5c4ca8-734c-11ef-b5ad-9577e7a7de79&p=3Dfd6c50ee-7739-= 11ef-a98a-c12e9d91840d&pt=3Dcampaign&t=3D1726838822&s=3D93fe2c338de52a92dfa= af295d6fb71f70965da2c91b02fcd9500e483b4b28812 [4] https://get.incogni.io/= aff_c?offer_id=3D1151&aff_id=3D16286 [5] https://techcrunch.com/2024/09/1= 7/snap-is-introducing-an-ai-video-generation-tool-for-creators/?utm_source= =3Dtldrai [6] https://www.theverge.com/2024/9/19/24249206/apple-intellige= nce-ios-18-1-public-beta?utm_source=3Dtldrai [7] https://techcrunch.com/2= 024/09/19/cruise-avs-return-to-bay-area-year-after-pedestrian-crash/?utm_so= urce=3Dtldrai [8] https://arxiv.org/abs/2402.06457?utm_source=3Dtldrai = [9] https://arxiv.org/abs/2409.12193v1?utm_source=3Dtldrai [10] https://g= ithub.com/chaoqi31/physmamba?utm_source=3Dtldrai [11] https://world.aiacc= eleratorinstitute.com/location/caioboston/?utm_source=3Dtldrai [12] https= ://world.aiacceleratorinstitute.com/location/caioboston/ [13] https://wor= ld.aiacceleratorinstitute.com/location/boston/ [14] https://github.com/Uc= as-HaoranWei/GOT-OCR2.0?utm_source=3Dtldrai [15] https://github.com/fisha= udio/fish-speech?utm_source=3Dtldrai [16] https://github.com/1x-technolog= ies/1xgpt/tree/main/genie?utm_source=3Dtldrai [17] https://futurism.com/o= penai-chatgpt-initiating-conversations?utm_source=3Dtldrai [18] https://m= istral.ai/news/pixtral-12b/?utm_source=3Dtldrai [19] https://www.oneusefu= lthing.org/p/scaling-the-state-of-play-in-ai?utm_source=3Dtldrai [20] htt= ps://www.ycombinator.com/companies/overlap?utm_source=3Dtldrai [21] https= ://techcrunch.com/2024/09/17/mistral-launches-a-free-tier-for-developers-to= -test-its-ai-models/?utm_source=3Dtldrai [22] https://github.com/orionw/p= romptriever?utm_source=3Dtldrai [23] https://refer.tldr.tech/21532aea/2= [24] https://hub.sparklp.co/sub_a89cbcf98f89/2 [25] https://advertise.= tldr.tech/?utm_source=3Dtldrai&utm_medium=3Dnewsletter&utm_campaign=3Dadver= tisecta [26] https://a.tldrnewsletter.com/unsubscribe?ep=3D1&l=3Deedf6b14= -3de3-11ed-9a32-0241b9615763&lc=3Ddf5c4ca8-734c-11ef-b5ad-9577e7a7de79&p=3D= fd6c50ee-7739-11ef-a98a-c12e9d91840d&pt=3Dcampaign&pv=3D4&spa=3D1726837268&= t=3D1726838822&s=3D4b3cdc6f756fd040ec3c8902cfb75b39e56069990f99dd162951f4c2= ce98323d [27] https://tldr.tech/ai/manage?email=3Dtldr%40synchro.net --OADsCkH_ Content-Type: text/html; charset=utf-8 Content-Transfer-Encoding: quoted-printable TLDR AI
Apple has released public betas of iOS 18.1, iPadOS 18.1, and macOS Sequ= oia 15.1 that feature new Apple Intelligence tools like text rewriting = ;

Sign Up |Advertise|<= a href=3D"https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fa.tldrnewslet= ter.com%2Fweb-version%3Fep=3D1%26lc=3Ddf5c4ca8-734c-11ef-b5ad-9577e7a7de79%= 26p=3Dfd6c50ee-7739-11ef-a98a-c12e9d91840d%26pt=3Dcampaign%26t=3D1726838822= %26s=3D93fe2c338de52a92dfaaf295d6fb71f70965da2c91b02fcd9500e483b4b28812/1/0= 10001920f9d5802-54728a72-39b3-44d5-8d78-b4aa8ce96145-000000/aSOjgRXVT9VJCS7= GNQlleURpET8--DDNzP0aIurLJ3M=3D371">View Online
TLDR

Together With 3D"Inco=

TLDR AI 2024-09-20

Your online privacy matters= .. Take back control with Incogni (Sponsor)

If you don't mind having your personal = data available to every spammer, scammer, and bad actor who's willing to pa= y for it, skip this ad.

Still here? Check out Incogni =E2= =80=94 it's the hassle-free way to protect your data privacy:

  • Incogni scans people search sites for your personal information and sen= ds removal requests on your behalf.
  • Within =C2=B114 days, your records are off the dark corners of the inte= rnet.
  • Every 10 days, Incogni does it all over again.
  • You stay in the loop with regular privacy reports.

Take back control. Reduce spam, scam, and cyber risk.

G= et 60% off Incogni with code TLDRAI (30 day money back guarantee= )

=F0=9F=9A=80

Headlines & Launches

Snap is introducing an AI v= ideo-generation tool for creators (2 minute read)

Snapchat has announced a new AI video-g= eneration tool for select creators that enables video creation from text an= d soon image prompts. The tool, powered by Snap's foundational video models= , will be available in beta on the web. Snap aims to compete with companies= like OpenAI and Adobe but has not shared output examples yet.
Apple Intelligence is now a= vailable in public betas (2 minute read)

Apple has released public betas of iOS = 18.1, iPadOS 18.1, and macOS Sequoia 15.1 that feature new Apple Intelligen= ce tools like text rewriting and photo cleanup. Only the iPhone 15 Pro, iPh= one 16, iPhone 16 Pro, and M1 iPads and Macs support these AI features. Fin= al versions are expected in October.
Cruise robotaxis return to = the Bay Area nearly one year after pedestrian crash (2 minute read)

Cruise is resuming operations in Sunnyv= ale and Mountain View, with human-driven vehicles for mapping and plans to = progress to supervised AV testing later this fall. This follows a settlemen= t and leadership change after an October 2023 crash. Cruise has issued soft= ware updates and signed a partnership with Uber for robotaxi services start= ing in 2025.
=F0=9F= =A7=A0

Research & Innovation

V-STaR: Training Verifiers = for Self-Taught Reasoners (31 minute read)

V-STaR is a novel approach to improving= large language models that utilizes both correct and incorrect solutions g= enerated during self-improvement to train a verifier, which then selects th= e best solution at inference time. The method has shown significant improve= ments in accuracy on code generation and math reasoning benchmarks compared= to existing approaches, potentially offering a more efficient way to enhan= ce LLM performance.
Fast 3D Generation from Sin= gle Images (31 minute read)

Vista3D is a new framework that generat= es 3D models from a single image in just 5 minutes. Using a two-phase appro= ach, it quickly forms rough geometry before refining the details, capturing= both visible and hidden aspects of objects for more complete 3D reconstruc= tions.
Heart Monitoring from Facia= l Videos (GitHub Repo)

PhysMamba is a new framework designed f= or remote heart monitoring via facial videos, addressing challenges in capt= uring long-range physiological signals.
=F0=9F= =A7=91=E2=80=8D=F0=9F=92=BB

Engineering & Resources

AIAI Boston: the East Coast= 's most significant summit for applied AI's builders & execs. =F0=9F=9A= =80 (Sponsor)

Uniting engineering teams & tech le= adership unleashing the LLM revolution, AIAI Boston returns on October 16-1= 8.

3 co-located summits. 500+ attendees. CXO speakers from Runway,= NVIDIA, Takeda, Optum.

Leaders =E2=9E=A1=EF=B8=8F <= a class=3D"c-link" href=3D"https://tracking.tldrnewsletter.com/CL0/https:%2= F%2Fworld.aiacceleratorinstitute.com%2Flocation%2Fcaioboston%2F/1/010001920= f9d5802-54728a72-39b3-44d5-8d78-b4aa8ce96145-000000/LtrxsQMCNymnjJB5ClVDhQL= uoTUfLI9u44fbH0pMvY0=3D371" rel=3D"noopener noreferrer" target=3D"_blank"><= span>apply for your Chief AI Officer Summit pass.

Engineers =E2=9E=A1=EF=B8=8F explore Gen= erative AI Summit & Computer Vision Summit.

GOT OCR (GitHub Repo)

A somewhat amazing advancement in gener= al-purpose optical character recognition (OCR) that can read text from imag= es with great performance. This particular version dramatically improves in= -the-wild OCR as well.
Fish Speech (GitHub Repo)

Powerful voice generation and single-sh= ot voice cloning. Completely open source and easy to get running.
1X Genie (GitHub Repo)

Genie is a video generation for world m= odel systems. 1x Robotics has open-sourced a version that mirrors the one i= t trained internally.
=F0=9F= =8E=81

Miscellaneous

<= /div>
OpenAI Says It's Fixed Issu= e Where ChatGPT Appeared to Be Messaging Users Unprompted (3 minute read)

A Reddit user reported that OpenAI's Ch= atGPT initiated a conversation unprompted, leading to speculation about new= engagement features. OpenAI acknowledged the issue and issued a fix, attri= buting it to a glitch with unsent messages. Debate continues over the authe= nticity of the incident, with similar reports from other users.
Announcing Pixtral 12B (8 m= inute read)

Pixtral 12B excels in multimodal tasks,= maintaining state-of-the-art performance on text-only benchmarks, and supp= orts variable image sizes in a 128K token context window. Its architecture = includes a new 400M parameter vision encoder and a 12B parameter multimodal= decoder based on Mistral Nemo. Pixtral outperforms many open and closed mo= dels in multimodal reasoning and instruction following without compromising= on text capabilities.
Scaling: The State of Play = in AI (13 minute read)

LLMs like ChatGPT and Gemini are becomi= ng increasingly capable as they scale up in size, data, and computing power= , leading to improved performance across various tasks. Current Gen2 models= like GPT-4 and Claude 3.5 are leading the market, with upcoming Gen3 model= s expected to further escalate capabilities and costs. The discovery of a n= ew scaling law in AI, pertaining to increased "thinking" during inference, = promises further advancements in AI performance beyond just model training.
=E2=9A= =A1

Quick Links

Overlap (Product Launch)

Overlap (YC S24) is a new AI-powered iO= S app that curates the best short video clips on literally any topic you're= interested in - built for those quick work or study breaks.
Mistral launches a free tie= r for developers to test its AI models (2 minute read)

Mistral AI has launched a free tier to = let developers fine-tune and build test apps with its models and slashed AP= I prices by over 50%.
A Promptable Retrieval Mode= l (GitHub Repo)

Promptriever is the first retrieval mod= el that can be prompted like a language model.

Love TLDR? Tell your friends and get rewards!

Share your referral link below with friends to get free TLDR swag!
https://refer.tldr.tech/21532aea/2
Trac= k your referrals here.

Want to advertise in TLDR? =F0=9F=93=B0

If your company is interested in reaching an audience of AI professionals a= nd decision makers, you may want to advertise with us.

If you have any comments or feedback, just respond to this email!

Thanks for reading,
Andrew Tan & Andrew Carr


If you don't want to receive fu= ture editions of TLDR AI, please unsubscri= be from TLDR AI or manage all of your TLDR newsletter subscripti= ons.
3D"" --OADsCkH_-- --- ï¿­ Synchronet ï¿­ Vertrauen ï¿­ Home of Synchronet ï¿­ [vert/cvs/bbs].synchro.net .