William W Collins

Pro-tip for those attempting to productionize gen-ai services based on text input and output: scrutinize the text input severely and relate that to the precise wording in the prompt. it is incredible *how* subjective text evaluation really is and I have learned just how *good* but subjective LLMs are. I find myself being confused about who's right in the evaluation: myself or the LLM! One thing stands out though: when I go back and read my prompt I often realize that the wording I used is actually not that clear, even though in everyday parlance another speaker *would seemingly know exactly what I was asking or saying*. It's a wild world.

2

California Governor Newsom vetoes the controversial AI Safety Bill! And the reasons behind it are worth noting. Stifling innovation is a major concern but so is the lack of nuance in the assessment of risk He states that the bill would 'give the public a false sense of security about controlling this fast-moving technology' Which honestly- is spot on! The Bill tries but both goes too far & not far enough. We covered the topic in great detail in our inaugural episode of TheOutpost.ai podcast Do check out our short of it here & let us know how you think the world will react to the veto! https://lnkd.in/gfs3vVN4

22

As you may be aware Meta's #Llama3 has been released. The best / most succinct overview I've seen of the release is by Andrej Karpathy on Twitter: https://lnkd.in/eH-2qbK5 Noting that the sequence length has been increased to 8,192 tokens, which is still relatively small compared to modern standards (e.g., GPT-4 at 128K tokens). And, while the inclusion of non-English tokens is a positive step, the 5% ratio is still quite low, making Llama 3 primarily an English-focused model. #ai #enterpriseai #llm

4

So awesome to see the AR team at Meta Reality Labs launch Orion glasses - their first version of stereoscopic Augmented Reality glasses and a true technological marvel. Congratulations to those everyone who pushed it over the finish line. While I was at Meta XR, I was fortunate to partner with the AR team heavily on the early prototypes of the product experience & technology. It was for a ground-breaking futuristic calling modality that seems to have not made it in this early version, but I am glad Mark hints it in his video here (https://lnkd.in/grna5EBR) at 1:30 mark - "Holograms ... where we are having this conversation in the future ... I am sitting on your couch next to you as a hologram ... we are just there and have this full sense of presence ...". If Meta continues to work on that vision, the future is going to be awesome! So you can imagine a world where I don't call my parents on my mobile phone with their video constrained in a 2D screen, but they can join me for chai on my dining table as a hologram, and vice versa. Where grandparents enjoy interacting with their grandkids playing in their living room in front of them through their Orion glasses as 3D holograms, not just viewing their video on a 6-inch phone. If you want to hang out with a close friend or significant other for fun to provide emotional support, you just hang out over a Holographic call right in each other's physical spaces.  I look back fondly on my time working towards that vision - working on user-research, designing how this novel consumer experience might work, and providing a high quality computer vision video calling experience within the constraints of gaze, brightness, thermals and wireless networks was one of the most intellectually stimulating work I have experienced. I can't wait for the day when the successor of Orion will truly defy distance and make the vision described above possible. Before 2030 perhaps?

27

3 Comments

So much agree with this. I see opposition to AI regulation and safety from some people, but as stated in this article, haven’t seen any elaboration on how it would be bad. It’s like saying that “don’t have a approval process for drug development as it would stifle innovation”. Or apply that to level 4 driverless cars. Folks who think that AI models and systems should not be looked at in the same vein, I would just point you to crypto disasters from last couple of years and how people thought then too that there shouldn’t be any barriers. If you’ve to fight the regulation, challenge the specific clauses rather than diss the whole thing. If you’re in tech, just talk to your friendly CISO and they would tell you what would happen if your company didn’t do those compliance certifications. https://lnkd.in/eJpv7EEe

9

In a recent talk, Eric Schmidt lists three #AI features that he believes will have unprecedented impact in the next 1-2 years. 💭 Insanely large context windows gives better short-term working memory. 📚 Self-learning AI agents self-improve by taking their learning, testing it, and then integrating the test feedback. ⚡️Through text-to-action the whole world gets a programmer at their service. Programmers are no longer bottlenecks for turning idea into reality.

10

1 Comment

🚨 **California AI Safety Bill Vetoed: What’s Next?** 🚨 I recently came across the news that Governor Newsome has vetoed the California AI Safety Bill. While the bill didn't make it into law, it's sparked a lot of discussions about AI safety regulations and their importance. This situation signifies deep concerns about the lack of structured AI safety measures and what it could mean for our industry. As someone who’s been closely following AI developments, I can’t help but reflect on the implications of this decision. While many tech companies are racing to innovate, the absence of regulatory guidance leaves us in a murky area where ethical frameworks and safety standards aren't established. This could lead to increased risks—think bias in AI algorithms, data privacy issues, and even potential misuse of AI technologies. On the flip side, the lack of regulations might create opportunities for innovative solutions and frameworks to emerge. Companies that proactively address safety and ethics in AI can differentiate themselves and position as leaders in responsible AI deployment. Firms that embrace these challenges could rise above competitors who may neglect these pressing issues. In a rapidly changing landscape, it’s vital for us as IT professionals to engage in these conversations. How can we balance innovation with safety? What frameworks can we build to ensure AI is developed and deployed ethically? I invite you all to share your viewpoints! How do you think the veto of the California AI Safety Bill will impact our industry? What steps are you considering to adapt to these changes? Let’s discuss! 💬💡 #AI #Technology #Regulation #Innovation #AIethics https://buff.ly/3XWTXPU #AISafety #CaliforniaPolitics #RegulationDebate #TechEthics

1

Anthropic's latest Claude AI model is taking human-AI interaction to a whole new level! 🖥️🤖 The new Claude 3.5 Sonnet can now use computers just like a human would - by looking at the screen, moving the cursor, clicking buttons, and typing text. This is a major leap forward in AI capabilities. Key features: Operates based on screenshots, emulating human computer use Can perform complex multi-step tasks across different applications Self-corrects and retries when encountering obstacles2 Accessible via API for developers to build with Why it's game-changing: This opens up incredible possibilities for AI assistance in data entry, form filling, software testing, and automating repetitive computer tasks13. Imagine having an AI assistant that can seamlessly work alongside you, handling tedious computer work while you focus on higher-level tasks. While still in beta, this technology demonstrates how AI is rapidly advancing toward more human-like interaction with digital interfaces. As it develops further, it could revolutionize how we work with computers and boost productivity across industries. The future of human-AI collaboration just got a lot more exciting! What do you think about this new capability? #AIInnovation #FutureOfWork Here is a video about it: https://lnkd.in/gh8aNYEu

6

With changing regulations regarding accessibility and a continued interest in human centered design, this article highlights the need for both. The guidance calls for a methodical, low risk approach as the organization matures in these practices. Great advice for all! https://lnkd.in/eFeD2ePw

18

Here is another quick read from Marty Cagan. Is Product Art or Science. This would translate to Are Product Managers Artists or Scientists? Do share your thoughts in comments below 👇

2

"A company is ultimately responsible for the things their representatives are saying." and "A robot can be a representative of a company.” Saw this quote from Katie Hass from the Utah Division of Consumer Protection as part of an article on if companies will be liable for what their AI agents (bots) say. This seems to come up every time a new technology comes up and different organizations now have the opportunity to interact with customers. It reminds me of articles where car dealers claimed that their on-site sales agents didn't have to honor prices posted to their websites for the same vehicle. While that situation and many others are long settled, it is interesting how the same principle comes up over and over -- and how the answers are (unsurprisingly) consistent about a brand's responsibility there. We're all faced with disruptions which challenge the existing way of working and our carefully crafted controls that manage risk. New tech, like AI Agents, creates these situations - the winners will be those that can help businesses successfully adopt the new models quickly while mitigating the risks to an acceptable level for different industries.

18

1 Comment

hype-o-ware (n): A sub-genre of vaporware, distinguished by conformance with current tech hype cycles. e.g., "We'll solve human aging via crypto" or "AI mumble mumble".

73

6 Comments

I will always side with open source, but I think we don’t have to worry about it that much since the machine learning community has been very forth coming in advances with algorithms and research either by arxiv papers or code examples even though the billionaire execs probably want to keep all these secrets to their companies the normal income ‘nerds’ that actually come up with this stuff publish and distribute their ideas, it’s shared and figured out by other nerds … Anyone can read Claude Shannon’s paper from 70 years ago to learn about entropy it’s not locked away behind some open ai paywall :) As compute gets cheaper it will enable more to train their own models. Plus tech employees are always moving around when they leave company X they take their methods with them and this spreads out the ideas and technology as well. It’s going to be difficult to try to regulate who does what with how ubiquitous the technology will be.

1

We live in a data-driven world today, and most enterprises are sitting on a goldmine of information (or even raw data). But like with crude oil, this raw data needs refinement to be truly valuable. Traditional databases (Sqls and NoSQLs), while essential, often struggle to capture the intricate relationships and interconnectedness that lie beneath the surface. This is where graph databases shine, offering a powerful tool to unlock the hidden potential of your enterprise's data. So, why knowledge graphs? Imagine your data as a vast network of interconnected nodes, each representing a piece of information. Graph databases excel at modeling these relationships, allowing you to explore and analyze data in ways that traditional databases simply can't. What about the problem of Disparate/Siloed Systems? One of the biggest challenges facing enterprises is data silos. Information is scattered across different systems (departments), making it difficult to get a complete picture (refer to Conway’s law). Graph databases act as a bridge, connecting these disparate systems and creating a unified view of your data. This interoperability enables you to gain deeper insights and make more informed decisions. Some context on Conway’s law: Conway's Law is a principle in software engineering that states: "Organizations which design systems are constrained to produce designs which are copies of the communication structures of these organizations."   In simpler terms, the way a company is structured and how its teams communicate will directly influence the structure and design of the software systems they create. If a company has silos or communication barriers, the software they produce will likely reflect those silos.   How do Knowledge Graphs support LLMs? For LLms to be useful, they rely on context to provide meaningful results. Graph databases provide the rich context that language models need to understand and respond to queries effectively (Read more about Retrieval Augmented Generation-RAG). By capturing the complex relationships between data points, graph databases help AI systems to be more accurate. Business never sleeps, so how do Knowledge Graphs adapt? As your business evolves, so too will your data. Traditional databases can be rigid, requiring significant effort to accommodate changes. Graph databases, on the other hand, are designed to be flexible. They can easily adapt to new relationships and data structures, ensuring that your technology stack remains relevant as your business grows. Bottom line! Graph databases offer a powerful solution to the challenges of data management in today's enterprises. By connecting disparate systems, providing rich context for AI, and offering unparalleled flexibility, graph databases enable organizations to extract maximum value from their data. If you're looking to understand more on Knowledge Graphs and their practical applications for your enterprise, check out this article by #neo4j #AI

19

8 Comments

Well worth going through the paper or slides, super well explained.

18

Congratulations Eshwaran Krishnan and proud of the Notion recognition! Thoughts to actionable workflows is an interesting use case and will enhance the productivity, so eager to see the NotesDev Beta version ! My linkedIn friends share your comments for NotesDev and DM Eshwaran Krishnan for additional inputs to build a disruptive productivity tool !!!

25

3 Comments