RT @TechstrongTV: Jonathan Siddharth, CEO and co-founder of the Palo Alto-based tech unicorn Turing, shares insights on how companies and #… - https://lnkd.in/gZ5hd-rM
The way humans reason does not exist on the internet.
AGI progress used to be constrained by compute and data.
Now, it’s human intelligence.
This is why Turing exists.
We train AGI and build with AI.
https://t.co/WOyxdvZZtF
Great interview with John Schulman, OpenAI co-founder and creator of ChatGPT on the road to AGI, RLHF, pre and post training.
John says, the models will get better as they train on harder tasks. They will start to do more complex tasks. They will go from code completions -> to doing entire projects, be able to understand a project at a high level, code it up, involving multiple files, iterate on tasks, test, deploy projects etc.
The internet/pre-training data doesn't have these complex reasoning chains to train on. The internet data is also commodity since all the models have access to the same and they've already ingested these "basic tokens". The "intelligent tokens" to take AGI to the next level, exist in the minds of software engineers and knowledge workers, doing complex projects in a real world setting.
This is why, we believe AGI progress used to be bottlenecked on compute and data. Now it's human intelligence.
What we need is "supervised human intelligence" to harness humanity's intelligence at scale, to move model reasoning/performance forward on complex tasks across every type of knowledge work you can imagine. This also needs specialized software to scale up these post training processes.
This is what we're building at Turing in addition to building the enterprise applications that will use these models.
Exciting times!
AI has transitioned from needing data to needing harnessed human intelligence at scale.
The commodity tokens are already used up and all the leading models are already trained on them.
The next jump both for pre and post training will come from supervised human intelligence at scale from companies like Turing.
Today, Reddit, Inc. announced its Q2 Earnings; our second quarterly financial results since becoming a public company in March 2024. In Q2 we:
- Achieved strong user growth as Weekly Active Uniques grew to an all-time high of 342M (+57% YoY)
- Continued to invest in full-funnel capabilities for advertisers, with total revenue increasing to $281M (+54% YoY)
- Launched Dynamic Product Ads, a significant milestone in our direct response ads business
- Acquired Memorable AI, bringing new Gen-AI campaign creation tools to advertisers
- Launched a content partnership program with the NFL, NBA, MLB, PGA & NASCAR
As we look to the future, we’re excited to:
- Scale machine translation to more international markets
- Continue improving search on Reddit with new formats & landing pages
- Expand the ability for users to create & earn money on Reddit
Learn more at: https://lnkd.in/gk2NuBVH
100% agree with Joanne ChenJaya G.Ashu Garg on this.
The future of work is "Service as software". That software will be AGI. AGI has the potential to automate knowledge work across all functions and all industries, to varying degrees.
I think the evolution will be,
co-pilots -> agents with human supervision -> agents (we don't have to go all the way in every cell in the function X industry matrix).
To achieve this, the bottleneck for AGI progress used to be compute and data. Now, it's human intelligence.
We need human intelligence at scale for post training to unhobble AGI across a wide variety of functions (sales, marketing, software engineering etc) and industries (retail, healthcare, hitech etc) and we need the data feedback loop from AGI products/workflows/solutions being actually used in production to do the work.
Turing's proud to be working with the world's leading AI companies and startups working on improving model performance and then deploying products built on those models with enterprises.
Eventually, companies will become 100X cheaper to start and operate. We are entering the golden era of productivity for Enterprises and startups.
AGI will unleash the world's untapped human potential. The impact of the human productivity unlock on our progress as a civilization, is hard to fully internalize. It's going to be massive.
How much of the scope of human work can the Service-as-Software model improve, widen, or replace?
$4.6 trillion worth.
How did Jaya and I calculate this?
We analyzed AI's capabilities in functions like marketing, sales, coding, and cybersecurity.
We also considered the costs companies incur for hiring people to perform these tasks and outsourcing general IT and business services.
Our findings indicate that AI and automation will significantly influence these and other functions, representing a massive $4.6 trillion opportunity within the next five years.
Swipe for the breakdown and check the comments for a link to the full article.
Foundation Capital
Co-founder & CEO, Virtualness, GenAI+blockchain platform to authenticate “moments that matter” | Board Director | Obsessed about enabling diverse leaders
Turing Presents AGI Icons: Rapid AI Advancement with Adam D'Angelo, CEO, Quora, and Board Member, OpenAI.
Inclusive of a pitch competition with the winner getting $250K - Congrats, Anish Agarwal, Interactionlabs.ai
Fascinating to see Turing’s evolution over the last six years since I have known Jonathan Siddharth. I first heard about Jonathan from Ashu Garg, Foundation Capital, and Kelly Graziadei as they wrote some of the first checks into Turing.
From Turing being the source for highly-vetted development talent from across the world to now enabling the worlds largest companies to advance and deploy AGI and powering the worlds leading models - it’s such a remarkable journey 🙌🏼
Don’t miss the last picture in the post taken in Turing office 6 years ago 😱. FYI some of Virtualness incredible early talent came from Turing. Saurabh Doshi Doshi
The conversation with Adam was so riveting that I forgot to take pictures of the beautifully moderated session 🤷♀️ 😊. Topics included how Gen AI is being used in Quora (a lot!), to perspectives on curriculum additions to get the next-gen ready for future jobs, to what each one of us should strive to do (be an expert in an LLM), to views on UGC vs GenAI content and more.
Many ideas sparked by conversations w fellow founders, investors and members from the ecosystem.
👏Congrats again Anish Agarwal for winning the pitch competition!
👏Congrats Jonathan Siddharth re the momentum and impact.
👏Congrats Ashu Garg on 16 years of Venture Capital — and the legacy of category defining companies you have backed as their first check!
#AGI#founderstories#TuringAGIIcons
There were 3 takeaways for me after the Turing event yesterday where Anita Ramaswamy drove an insightful discussion with Adam D'Angelo and Jonathan Siddharth -
- We are still in the honeymoon phase with genAI. Too many promises and dreams. But the GTM strategies need to be proven out. Large companies like OpenAI, Meta can bet big because they will direct the future of this space. The rate of real Enterprise adoption of AI tools in the next 12 months will determine whether the hype is translating to dollars.
- We should NOT be chasing AGI because there isn’t a uniform definition of what it is. This was clear when the speakers had different perspectives about what they think it was. Every member of the audience had their own opinion too.
- Young engineers and college grads in the CS area will need more support/training to be successful in an AI first world. This is a conversation that is urgent so we don’t lose momentum and sideline emerging talent.
Also, congrats to Anish Agarwal and Interaction Labs (https://lnkd.in/gXwDTZNP) for their win in the “Ideas that are production ready” contest.
#genai#artificialintelligence#agi#aitraining