UNGA79 and the ‘Pact for the Future’

The ink on the Pact for the Future was barely dry

a global handshake

for the unpredictable rhythms of modernity.

The ‘Pact for the Future,’ adopted at the Summit of the Future on 22 September 2024, sets out an ambitious agenda to address climate change, digital transformation, and peace while fostering agile global governance. 

Gold has soared to a record high of $2,629 per ounce following the US Federal Reserve’s recent interest rate cut.

AI models, including ChatGPT and Cohere, once depended on low-cost workers for basic fact-checking. Today, these models require human trainers with specialised knowledge in medicine, finance, and quantum physics. 

While South Korean memory giants Samsung Electronics and SK hynix experienced a significant sales increase in China during the first half of this year, the report by the Korea Eximbank Overseas Economic Research Institute indicates that South Korea’s reliance on China for critical semiconductor raw materials is also growing.

a lighthouse – navigating

Moscow’s efforts have focused on supporting Donald Trump and undermining Kamala Harris.

to leap from the past into an uncertain, but ambitious, tomorrow.

Fighterbomber

was lost due to the absence of recordings, transcripts, or even note taking.

Everyone makes hundreds of decisions a day, but most of the time they aren’t conscious of the decisions they are making.

The first is the geostrategic importance of technology – in particular data consuming and content producing technologies (which explains why the US has been so keen to limit the power of China’s TikTok).

I suspect for instance there will be a lot more civil and corporate legal cases to establish the ownership of data-sets and the access to them by AI engines, as is the case between the New York Times and OpenAI.

technology weighs on many public policy debates now – such as the access to mobile phones in schools, the demands that data centres place on electricity grids and the TikTok-ification of politics.

Productive stupidity means being ignorant by choice. Focusing on important questions puts us in the awkward position of being ignorant. One of the beautiful things about science is that it allows us to bumble along, getting it wrong time after time, and feel perfectly fine as long as we learn something each time. No doubt, this can be difficult for students who are accustomed to getting the answers right. No doubt, reasonable levels of confidence and emotional resilience help, but I think scientific education might do more to ease what is a very big transition: from learning what other people once discovered to making your own discoveries. The more comfortable we become with being stupid, the deeper we will wade into the unknown and the more likely we are to make big discoveries.

Like other comments, I was also initially surprised. But I think the gains are both real and easy to understand where the improvements are coming from.

Chain-of-Thought and Self-Reflection are also very simple, and this was built with a SYNTHETIC dataset. Can you imagine if we used people to carefully curate a much better dataset? If we experimented with other forms of CoT, self-reflection, multi-agent reasoning, etc? There are so many things to try, and you don't have to be at OpenAI to try them!

Imagine a text chunk including a sentence like “Donald Trump says that there should be a wall on the southern border of the US.” The meaning of this sentence could not be understood unless one knows the broader context—such as if Donald Trump was incumbent when he made such a statement. If he was, then the sentence probably conveyed the official stance of the US at that time; otherwise, it would be Mr. Trump’s personal stance. The Manager argued that an RM model—such as the one used by the lab at that time—only conducted semantic searches and could not capture context beyond the text chunk; it might mistake official country stances, retrieve the wrong paragraphs for the chatbot, and lead the user to believe in wrongful answers.

The way that the AI reporting system should transform information into knowledge now includes comparing current knowledge to historical knowledge,

It turned out that, to her, the most valuable information would not be what was said this year but what was said this year that was different from all previous years.

The AI lab brought in a linguist, who broke down conversations into questions and answers; the linguist taught the AI lab various question types one could pose in a conversation and ways to detect which type they were (rhetorical, open, etc.). The next step was to understand the answers; the AI lab consulted a debater who framed responses in terms of arguments, which were then understood as key points with corresponding supporting facts. Taking in these lessons, the AI lab instructed the reporting system to take each speaker’s paragraphs from a transcript, extract key points and supporting evidence, and present a session summary in this format.

The course team prepares various courses that the Foundation delivers to diplomats or higher-education students. The reporting team takes care of the daily monitoring and updating of digital news and global policy trends on the Foundation’s website; it is also the team that used to provide live coverage of major international events, such as the UN General Assembly (UNGA) and Internet Governance Forum (IGF). The creative lab designs social media campaigns and visuals for all published material. The tech team manages the technical infrastructure, from websites to internal tools and applications, that allows the Foundation to function.

AI models are not merely machines that solve our problems; they are our imagination of the world and its problems; they are our respective worldviews, personalities, ambitions, and desires reformulated, reconfigured, and translated into codes.

In the Director’s envisioned knowledge management system, the technology product must not be a mere standalone tool but a part of a workflow that processes information

The AI reporting system takes an audio-visual recording of a session, transcribes all speakers’ speeches via a transcription model (TM), and then generates summaries of various formats and knowledge graphs based on the transcripts via several summarization models (SMs) and a knowledge-graph generating model (KGGM).

We recognize the importance of  continuing to pursue governance reforms

Trade provides another example of the failure of multilateral negotiations (Action 5). The WTO (which has replaced UN Trade and Development, or UNCTAD, in this function) has rules that are being willfully ignored. The US has knee-capped the dispute mechanism by refusing to allow the appointment of judges; the world’s largest exporter, China, continues to claim developing-country status and hides information on its trade policies. Global trade negotiations began in 2001 in Doha and died in 2015. Determined multilateralists are calling for WTO reform but there is no sign that the changes will happen. There are echoes here of the ill-fated efforts to change the Security Council.

But the COPs have been singularly ineffective for several reasons. One is the UN secretariat’s timidity. It has been far too indulgent to member states in allowing voluntary national targets to reduce carbon emissions. It has permitted petro-states to host COP meetings, like the United Arab Emirates last year and Azerbaijan this year, and it has failed to counter the advocacy by oil and gas interests. Progress is measured by slight adjustments of language in its outcome statements while downplaying actual outcomes in CO2 emissions.

the First UN of member states — and particularly the veto wielding P3 of Russia, China and the United States — and the Second UN of secretariats.

Respect for human rights, self-determination, territorial integrity, freedom of navigation, democratic governance, free movement of goods, economic openness etc. are important values that may be invoked by a court to assist it in the interpretation of legal rules but they are not rules of law as commonly understood.

the narrow positivist perception of international law to include soft law, including the standards and recommendations of international standard-setting organizationsFootnote 17 and conferences and rules made by non-state actors.

Linguistic research had identified 80,000 “noun synonym sets”, or synsets: groups of synonyms that described the same sort of thing.

calls the tech giant’s models “open weights”, rather than open source.

On one side are the open-source purists in favour of decentralised “little tech”. On the other are closed-source realists who argue that greater centralisation and control are better for safety and national security.

Starlink, a satellite-internet firm, could be a stopgap, but regulators have blocked it in at least seven countries including South Africa.

Mobile-phone services directly contribute $170bn to Africa’s gdp.

We’re testing SearchGPT, a temporary prototype of new AI search features that give you fast and timely answers with clear and relevant sources.

Impressive AI video generator KLING now available as web version

a paper introducing OpenDevin, a framework for developing AI agents that interact with environments similar to human programmers.

It’s OpenAI, Google, and Anthropic vs. Meta, Llama, and xAI. One thing is for certain: open-source big AI is going to be competitive.

A 2009 agreement insisted on by the European Commission meant that Microsoft could not make security changes that would have blocked the update from cybersecurity firm Crowdstrike that caused an estimated 8.5 million computers to fail, the Big Tech giant said in comments to the Wall Street Journal newspaper.

Modi sent his foreign minister, Subrahmanyam Jaishankar, to the Shanghai Cooperation Organisation (SCO) at its annual meeting in the Kazakhstan capital, Astana, which was also attended by Putin and the Chinese president, Xi Jinping.

“expert systems” which used symbolic logic to capture and apply the best of human know-how.

Amos Hochstein, Joe Biden’s envoy for energy security, hopes that, by redirecting exports to the Atlantic, supply chains will shift decisively westward.

the work

Superintelligence (also known as Artificial Super Intelligence, or ASI) is defined much more clearly as an AI having intelligence ‘far surpassing that of the brightest and most gifted human minds’.

likely

undefined marketing term

most people

simulating reasoning, avoiding hallucinations, and enhancing the capacity to use external tools.

Join our community of global citizens empowered by The New Humanitarian’s independent journalism from the heart of crises, conflicts, and disasters.

tested this idea by “spiking” (ie, turning up) a feature associated with the Golden Gate Bridge.

“mechanistic interpretability”.

“lots of different brains looking at the same source at the same time”.

I fully agree with the statement.  Water is a critical resource that underpins various sectors of the economy, including agriculture, manufacturing, energy, and cities

The New Popular Front (NFP) coalition led by Melenchon and its allies won a plurality in France’s National Assembly in the second round of the parliamentary elections on Sunday, taking 187 seats in the 577-seat National Assembly, interior ministry figures show.

negative effects of brokering on brokers and those whose well-being is affected by the activities of brokers

a digital factory can stitch together emerging technologies (AI, machine learning, and intelligent automation) with other tools, such as ServiceNow, Salesforce, and SAP, to replace or augment existing processes.

This approach led to a much more business and value-focused reporting capability, including:

Instead of rushing to use technology to automate existing reports, they asked fundamental questions about what reporting is needed, why, and how they could pivot this set of activities to growth while reducing costs.

LMSYS Chatbot Arena

This constituted 5% of the Indonesian population, a higher casualty rate than among many major belligerents, including the United States (0.3%), Britain (0.9%) and Japan (3.9%).

There is a lot of innovation theatre.”

For its part, the icrc warns that ai systems are potentially unpredictable, opaque and subject to bias, but accepts they “can facilitate faster and broader collection and analysis of available information…minimising risks for civilians”.

“If…machines produce a lower false positive and false negative rate than humans, particularly under pressure, it would be unethical not to delegate authority,”

Lavender is likely to be what experts call a decision-support system (dss)

Studies also indicate that llms preferentially cite other papers that are highly cited in a field, potentially reinforcing existing biases and limiting creativity

The Ultimate Guide to Video Aspect Ratio

Optimize for online viewing

New UNESCO report warns that Generative AI threatens Holocaust memory

In a more or less explicit way, this constitutive power dimension of technology always includes the worldview of those who invented and developed it.

the rules, norms and standards

“Kafkaesque”, first coined in the late 1930s, has become a label for nightmarish, complex and illogical situations, from government investigations to customer-service lines to nowhere.

Multiple Vector Databases — Connect to multiple vector databases to enhance Agent performance.

Even now, investors are lukewarm about such gambles. In recent weeks they have applauded Google’s capex plan

for as long as there

told the World Economic Forum

In a report released on April 5th, the Threat Intelligence team at Microsoft,

only four of the 11 countermeasures were widely endorsed in the research literature:

a technique known as “psychological inoculation” or “pre-bunking”.

the Global Engagement Centre, an agency in the us State Department.

The “news media bargaining code”, passed in 2021, obliged Google and Facebook to pay news outlets when search results or social-media feeds linked to their articles.

In future, whenever Meta faces demands for payment, “they will just walk away,” Mr Doub said.

News Media Canada, which supports the new law, is attacking Meta on another front, lodging a complaint with Canada’s competition regulator, arguing that Meta’s link-blocking is an abuse of its dominant position.

NewsGuard, a fact-checking organisation, found that “unreliable” sources accounted for 6.9% of engagement with news on Facebook in Canada in the 90 days after the blackout, compared with 2.2% in the 90 days before.

Energetic City has signed up 5,000 readers to an email newsletter, built a mobile app and put up posters around town with qr codes leading to its website. “We basically have to reteach our readers how to find the news,”

The law promised “fairness in the Canadian digital-news marketplace”.

How does the way the DSA addresses (and differentiates between) hate speech and harmful content impact Palestinian digital rights on major online platforms?

As a marketer, Jun was all about making quick decisions and fast iterations. Riley, however, followed the stringent protocols of their production team and was methodical and risk-averse.

Getting the Innovation Ecosystem Ready for AIAn IP policy toolkit

Evaluating Language Model Capabilities

The Baader–Meinhof phenomenon is a cognitive bias referring to the tendency to notice something more often after noticing it for the first time, leading to the belief that it has an increased frequency of occurrence.

Senior executives frequently are too far removed from day-to-day operations to understand what truly needs to change. Consequently, top-down solutions tend to be superficial or at least short-lived. Frontline managers, meanwhile, often lack the contextual understanding to challenge existing processes, and so trim around the edges rather than propose major changes. But midlevel executives tend to have enough experience to see the shortcomings in current operations—and aren’t so close to the ground that they get lost in the weeds.

notably, a lack of urgency, insufficient leadership, limited vision, poor communication, and a shortage of “quick wins.”

Having a crisp mental model around a problem, being able to break it down into steps that are tractable, perfect first-principle thinking, sometimes being prepared (and able to) debate a stubborn AI — these are the skills that will make a great engineer in the future, and likely the same consideration applies to many job categories.

the gold standard of learning — personalized tutoring — may be available to everyone.

the “rules-based international order”,

Rarely have international courts been busier.

could triple its share of American power consumption by 2030, to 7.5%.

“global capability centres”

more action-oriented discussions at the UN on international security issues in cyberspace

In an effort to increase digital solidarity in the realm of foreign assistance, USAID launched the Donor Principles for Human Rights in the Digital Age in partnership with Canada’s International Development Research Centre (IDRC),

action-oriented discussions

Satellite Communication Networks

The Future of AI Technologies Governance

Challenges of the Digital Economy

Cyber Attacks and National Security Threats

The rapid growth of AI technology, however, comes with the significant risk that its use may exacerbate inequality and economic instability, stifle competition, cause consumer harm, aggravate discrimination and bias, invade privacy, enhance malicious cyber activity, and improve authoritarian capabilities for surveillance and repression. AI will challenge how we compensate for the uses of intellectual property as well as authenticate, label, or detect synthetic content. AI may also require workforce adaptations across economies; the rising energy demands of high-end AI chips and data centers could become a significant barrier to developing local capabilities.

Digital solidarity is a willingness to work together on shared goals, to help partners build capacity, and to provide mutual support.[1] Digital solidarity recognizes that all who use digital technologies in a rights-respecting manner are more secure, resilient, self-determining, and prosperous when we work together to shape the international environment and innovate at the technological edge. Central to the tenets of digital solidarity are efforts to support allies and partners, especially emerging economies, to fully seize the opportunities presented by new technologies and sustainably pursue their economic and development goals. Digital solidarity aligns U.S. national interests with those of our international partners through compatible approaches to technology governance, sustains strong partnerships with civil society and the private sector, and embraces cybersecurity resilience built on a diversity of products and services made by trusted technology vendors. It highlights the mutual support that the United States and its partners offer one another to counter and respond to malicious cyber operations, cybercrime, and other digital harms, and promotes cooperative efforts among states and civic actors to defend and advance human rights. In addition, the concept of digital solidarity rests on efforts to build digital and cyber capacity so that partners are not only better able to build a defensible and resilient digital ecosystem over the long term but are also able to respond and recover quickly when incidents that threaten security, safety, and rights happen. The actions and efforts of this strategy are intended to demonstrate and build digital solidarity with partners across the globe.

the October 2022 National Security Strategy (NSS)

working together to offer mutual assistance to the victims of malicious cyber activity and other digital harms; assist partners – especially emerging economies – in deploying safe, secure, resilient, and sustainable technologies to advance their development goals; and builds strong and inclusive innovation economies that can shape our economic and technological future.

to build digital solidarity

a subtle difference in their meaning

The vocabulary of disinformation

First is “placement”, or the initial posting of the lie.

the Israeli government is split.

AidData

“true multilateralism” where “universal security” replaced “universal values”

It has built an index of states’ power from 1960 to 2022.

It is also claimed that the IDF would be willing to risk killing 15-20 civilians in order to strike a Hamas fighter. For Hamas battalion or brigade commanders, that number rose to more than 100 civilians.

But in 2016 Heather Roff and Richard Moyes, then writing for Article 36, a non-profit focused on the issue, cautioned that a person “simply pressing a ‘fire’ button in response to indications from a computer, without cognitive clarity or awareness”, does not meaningfully qualify as “human control”.

the eu ’s directive on digital copyright was passed in 2019, when generative AI was not a thing. “There is no way the Europeans would pass [such a directive] today,” says Mr Sag.

Tennessee last month passed the Ensuring Likeness Voice and Image Security ( ELVIS ) Act, banning unauthorised deepfakes in the state.

A number of news publishers have reached a similar conclusion

A survey of media executives in 56 countries by the Reuters Institute found that 48% expected there to be “very little” money from AI licensing deals.

Shutterstock, a photo library, has licensed its archive to both Open AI and Meta, the social-media empire that is pouring resources into AI . Reddit and Tumblr, online forums, are reportedly licensing their content to AI firms as well.

Partners so far include the Associated Press, Axel Springer (owner of Bild and Politico), Le Monde and Spain’s Prisa Media.

Outside America, the legal climate is mostly harsher for tech firms.

The lawyering is now happening.

tech firms are again seeking forgiveness rather than permission

Common Crawl,

the model learns by trial and error how tokens are normally combined.

“They have created an amazing edifice that’s built on a foundation of sand.”

AI s are trained on vast quantities of human-made work

In 2022 he proposed a framework called “Joint Embedding Predictive Architecture” ( JEPA ),

Andrej Karpathy, a researcher formerly at Open AI , explained in a recent talk that current LLM s are only capable of “system 1” thinking. In humans, this is the automatic mode of thought involved in snap decisions. In contrast, “system 2” thinking is slower, more conscious and involves iteration.

Since 2017 most AI models have used a type of neural-network architecture known as a transformer (the “ T ” in GPT ), which allowed them to establish relationships between bits of data that are far apart within a data set.

When the LLM is in its learning phase, it compares its predictions against the version of reality available in its training data. If these diverge, the algorithm makes small tweaks to each layer of the network to improve future predictions. That makes it computationally intensive and incremental.

Companies like Scale AI and Surge AI have built large networks of people to generate and annotate data, including P h D researchers solving problems in maths or biology. One executive at a leading AI startup estimates this is costing AI labs hundreds of millions of dollars per year.

Epoch AI , a research outfit, estimates the well of high-quality textual data on the public internet will run dry by 2026.